Re: access.log
--On Thursday, October 17, 2002 10:41 AM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Andy Elacion, Jr. schrieb: There are times that our subscriber would like to verify their transactions and we have hard time in doing this. We do it manually like grep the accesslog for a particular cell no and it give us a pain. Another thing is to monitor the reply of our system. hehe, we know about that ;) A MySQL support for logging may be added in the future. But to be honest I'm more interested in getting a Kannel module API done before embedding to much individual external software support into it. I was thinking of the same. Having such an API makes this all way more easy. Stip,e have you looked at that module API I once posted?? Harrie Internet Management Consulting mailto:harrie;mod-snmp.comhttp ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
RE: MMS Task Force
HI Alex, --On Thursday, September 12, 2002 5:02 PM +0100 Alex Judd [EMAIL PROTECTED] wrote: Agreed. The main priority should be to be complete the functionality needed to be able to send and recieve messages to and from a variety of different 1st phase released MMS-Cs as per the current live v1 WAP MMS protocol implementation. Since there were also some question like 'I don't know it all??' maybe it is wise to create a web page that provides links to specific (useful) documents. From this point of view adding complete support for SAR is one of the current blocks and we have a good start in this area with the recent submissions. Testing is obviously needed here and access to the amount of software/hardware you have would be great. Potentially I would also suggest a seperate cvs branch which can be developed on without breaking the rest of the functionality. As most of us here use the cvs version of Kannel as staging (or often as live), I'd like to make sure that the changes don't stop the bug fixes. I agree here. It was the way forward I see too. The branch also allows the group to experiment more. Stipe - any thoughts? Some time ago he mentioned he would be on vacation. That could be now, since he mostly reacts quickly. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: FW: query on kannel usage.
--On Friday, September 6, 2002 3:01 PM +0530 Anupama R [EMAIL PROTECTED] wrote: Hello, Posted this message to users list. Got no response :(( We are proposing to use the gateway for our product. Is this okay? Can anyone please reply? I suggest you read the file COPYING. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: Kannel automatic compilation test
--On Wednesday, September 4, 2002 10:10 AM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Stipe, Would this patch solve it. Not knowing the protocol spec of SMASI. I can try. But I took the code from smsc_smpp.c as template and AFAIK, they had no such mungy p = *(p); thingies :)) I came up with this, since it is already in some places of the SMASI code. Since I believe you wanted a NO-OP function, but this is mainly an attempt to read your mind. Stipe Harrie Hazewinkel schrieb: Stipe, Would this patch solve it. Not knowing the protocol spec of SMASI. Index: smasi_pdu.c === RCS file: /home/cvs/gateway/gw/smsc/smasi_pdu.c,v retrieving revision 1.2 diff -u -r1.2 smasi_pdu.c --- smasi_pdu.c 19 Aug 2002 09:15:16 - 1.2 +++ smasi_pdu.c 3 Sep 2002 12:16:21 - @@ -140,7 +140,7 @@ } switch (pdu-type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); # define COMATERMINATED(name) \ if (p-name != NULL) { octstr_append_cstr(os, #name); \ octstr_append_char(os, '='); \ @@ -177,8 +177,8 @@ type_name = (temp ? octstr_get_cstr(temp) : ); if (strcmp(type_name, dummy) == 0) type = 0; -#define NONTERMINATED(name) -#define COMATERMINATED(name) +#define NONTERMINATED(name) p = *(p); +#define COMATERMINATED(name) p = *(p); # define PDU(name, id, fields) \ else if (strcmp(type_name, #name) == 0) type = id; # include smasi_pdu.def @@ -191,7 +191,7 @@ skip_until_after_colon(data_without_len, pos); switch (type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); # define COMATERMINATED(name) \ if (octstr_str_compare(field_name, #name) == 0 field_value != NULL) \ p-name = octstr_duplicate(field_value); @@ -226,7 +226,7 @@ debug(sms.smasi, 0, SMASI PDU %p dump:, (void *) pdu); debug(sms.smasi, 0, type_name: %s, pdu-type_name); switch (pdu-type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); # define COMATERMINATED(name) \ octstr_dump_short(p-name, 2, #name); # define PDU(name, id, fields) \ Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: WSP TO HTTP SLOW !!
--On Wednesday, September 4, 2002 6:32 PM -0700 denzel [EMAIL PROTECTED] wrote: HI all ! We sent a large messege ( 16k ) from the phone to the kannel wap gateway (posting large message content to a url). The message was posted ok. But it took a lot time to complete ( around a minute ). Debug output shows that the bottleneck is with the wsp trying to write the content to http socket. Does anyone know why ? I can only guess, for instance a slow connection to the HTTP server, but if you provide more info such as the debug output you refer to it would help. Also, how quickly was the message send from the phone and received in the gateway, maybe that was the reason why it took a minute to write the complete message to the HTTP socket. cheers, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: Kannel Http SMSC
Hi, I am not sure if I understand it correctly. You have Kannel speaking to an SMSC with HTTP and now you want to place an HTTP proxy is in the middle?? Please view in a fixed-width font such as Monaco or Courier. +--+ ++ +-+ | | || | | |KANNEL| | PROXY | | SMSC HTTP | | +-++-+ | | | || | | | | || | | | | || | | +--+ ++ +-+ I would say as long the proxy allows HTTP request to be passed (and of course the network functions properly) and I am assuming that the proxy does not change the HTTP re quests. Kannel can reach the SMSC HTTP where it normally should have invoked its HTTP requests on the SMSC directly. Is this what you mean?? Harrie --On Tuesday, September 3, 2002 4:43 AM -0700 Atef lotfy [EMAIL PROTECTED] wrote: Dear Sirs, I have kannel development ver 1.1.6. I want to configure kannel SMSC to be an HTTP,however,Kannel lies behind a proxy. Are the HTTP SMSC http requests forwarded also from the proxy server configurations in the bearer box Or these configurations are only for http requests for SMS service? Please Advice, Atef Lotfy Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: Kannel automatic compilation test
Hi All, --On Tuesday, September 3, 2002 7:06 AM +0100 kannel [EMAIL PROTECTED] wrote: Kannel compilation test. Host: Linux geodude.3glab.org 2.4.2-2 #1 Sun Apr 8 20:41:30 EDT 2001 i686 unknown Kannel compilation had warnings or failed. the warnings on the unused variable are caused by an expansion of the 'define macros' into code like this. case 0x0002 : { struct LogonConf *p = pdu-u. LogonConf ; } break; which is on its turn a result of this code for the define: #define PDU(name, id, fields) \ case id: { struct name *p = pdu-u.name; fields } break; and this definition: PDU(LogonConf, 0x0002, HEADER ) where I believe the HEADER variable is defined as NONTERMINATED. The define of NONTERMNIATED is defined as #define NONTERMINATED(name) This goes for the warnings In file included from gw/smsc/smasi_pdu.c:151: gw/smsc/smasi_pdu.def: In function `smasi_pdu_pack': gw/smsc/smasi_pdu.def:32: warning: unused variable `p' gw/smsc/smasi_pdu.def:49: warning: unused variable `p' gw/smsc/smasi_pdu.def:143: warning: unused variable `p' gw/smsc/smasi_pdu.def:148: warning: unused variable `p' Stipe, hope this helps you to fix it, since you added this SMASI SMSC. Harrie Output of 'make -s': In file included from gw/smsc/smasi_pdu.c:151: gw/smsc/smasi_pdu.def: In function `smasi_pdu_pack': gw/smsc/smasi_pdu.def:32: warning: unused variable `p' gw/smsc/smasi_pdu.def:49: warning: unused variable `p' gw/smsc/smasi_pdu.def:143: warning: unused variable `p' gw/smsc/smasi_pdu.def:148: warning: unused variable `p' In file included from gw/smsc/smasi_pdu.c:213: gw/smsc/smasi_pdu.def: In function `smasi_pdu_unpack': gw/smsc/smasi_pdu.def:32: warning: unused variable `p' gw/smsc/smasi_pdu.def:49: warning: unused variable `p' gw/smsc/smasi_pdu.def:143: warning: unused variable `p' gw/smsc/smasi_pdu.def:148: warning: unused variable `p' In file included from gw/smsc/smasi_pdu.c:234: gw/smsc/smasi_pdu.def: In function `smasi_pdu_dump': gw/smsc/smasi_pdu.def:32: warning: unused variable `p' gw/smsc/smasi_pdu.def:49: warning: unused variable `p' gw/smsc/smasi_pdu.def:143: warning: unused variable `p' gw/smsc/smasi_pdu.def:148: warning: unused variable `p' gwlib/http.c: In function `http_send_reply': gwlib/http.c:1976: warning: unused variable `p' test/test_cfg.c: In function `main': test/test_cfg.c:6: warning: `ret' might be used uninitialized in this function --- Output of 'CFLAGS='-Wall -O2 -g' ./configure': creating cache ./config.cache checking cvs checkout date... 20020902 Configuring for Kannel gateway version cvs-20020902 ... Running system checks ... checking host system type... i686-pc-linux-gnu checking for gcc... gcc checking whether the C compiler (gcc -Wall -O2 -g ) works... yes checking whether the C compiler (gcc -Wall -O2 -g ) is a cross-compiler... no checking whether we are using GNU C... yes checking whether gcc accepts -g... yes checking for a BSD compatible install... /usr/bin/install -c checking for ranlib... ranlib checking for bison... bison -y checking for flex... flex checking for yywrap in -lfl... yes checking for convert... /usr/X11R6/bin/convert checking for perl... /usr/bin/perl checking for log in -lm... yes checking for accept in -lsocket... no checking for inet_ntoa in -lnsl... yes checking for inet_ntop in -lresolv... yes checking for inet_ntop in -lbind... no checking for pthread_exit in -lpthread... yes checking for pthread_exit in -lc_r... no checking how to run the C preprocessor... gcc -E checking for ANSI C header files... yes checking for sys/ioctl.h... yes checking for sys/time.h... yes checking for sys/types.h... yes checking for unistd.h... yes checking for sys/poll.h... yes checking for pthread.h... yes checking for getopt.h... yes checking for syslog.h... yes checking for gettimeofday... yes checking for select... yes checking for socket... yes checking for strdup... yes checking for getopt_long... yes checking for getopt... yes checking for socklen_t in sys/socket.h... yes checking for getopt in stdio.h... no checking for getopt in unistd.h... yes Checking POSIX threads support ... checking for working pthreads... no Checking for libxml2 support ... checking for xml2-config... /usr/bin/xml2-config checking libxml version... 2.3.11 Configuring DocBook support ... checking for jade... jade checking for jadetex... jadetex checking for pdfjadetex... pdfjadetex checking for dvips... dvips checking for fig2dev... fig2dev checking for convert... /usr/X11R6/bin/convert checking for /usr/lib/sgml/stylesheet/dsssl/docbook/nwalsh/html/docbook.dsl... no checking for /usr/lib/sgml/stylesheets/nwalsh-modular/html/docbook.dsl... yes checking for /usr/lib/sgml/stylesheet/dsssl/docbook/nwalsh/print/docbook.dsl... no checking for /usr/lib/sgml/stylesheets/nwalsh-modular/print/docbook.dsl... yes
Re: Kannel automatic compilation test
Stipe, Would this patch solve it. Not knowing the protocol spec of SMASI. Index: smasi_pdu.c === RCS file: /home/cvs/gateway/gw/smsc/smasi_pdu.c,v retrieving revision 1.2 diff -u -r1.2 smasi_pdu.c --- smasi_pdu.c 19 Aug 2002 09:15:16 - 1.2 +++ smasi_pdu.c 3 Sep 2002 12:16:21 - @@ -140,7 +140,7 @@ } switch (pdu-type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); #define COMATERMINATED(name) \ if (p-name != NULL) { octstr_append_cstr(os, #name); \ octstr_append_char(os, '='); \ @@ -177,8 +177,8 @@ type_name = (temp ? octstr_get_cstr(temp) : ); if (strcmp(type_name, dummy) == 0) type = 0; -#define NONTERMINATED(name) -#define COMATERMINATED(name) +#define NONTERMINATED(name) p = *(p); +#define COMATERMINATED(name) p = *(p); #define PDU(name, id, fields) \ else if (strcmp(type_name, #name) == 0) type = id; #include smasi_pdu.def @@ -191,7 +191,7 @@ skip_until_after_colon(data_without_len, pos); switch (type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); #define COMATERMINATED(name) \ if (octstr_str_compare(field_name, #name) == 0 field_value != NULL) \ p-name = octstr_duplicate(field_value); @@ -226,7 +226,7 @@ debug(sms.smasi, 0, SMASI PDU %p dump:, (void *) pdu); debug(sms.smasi, 0, type_name: %s, pdu-type_name); switch (pdu-type) { -#define NONTERMINATED(name) +#define NONTERMINATED(name) p = *(p); #define COMATERMINATED(name) \ octstr_dump_short(p-name, 2, #name); #define PDU(name, id, fields) \ Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: Kannel Http SMSC
--On Tuesday, September 3, 2002 5:17 AM -0700 Atef lotfy [EMAIL PROTECTED] wrote: Dear Sir, Thank you for your response. what I really want to do is to make kannel connects to an SMSC Http which is a web site (www.deltica.com) that has an http client interface to send SMS. kannel itself is behind an http proxy ( as in your drawing). The problem is that I have configured Kannel to use http proxy,but Kannel always fails to connect although the site and the proxy are working well. Kannel always gives me: connect failed system error 128: Network is unreachable error connecting to server at port 80 Can you help me fixing this ? I can try. So far it looks to me more a network problem were you cannot connect to the proxy on port 80 from the kannel system. Can you try to connect to port 80 on the proxy with the command: $ telnet proxy-hostname 80 Trying some IP... Connected to proxy-hostname. Escape character is '^]'. At this point you have a tcp connection to the proxy on port 80. 2 returns should close it for you. --- Harrie Hazewinkel [EMAIL PROTECTED] wrote: Hi, I am not sure if I understand it correctly. You have Kannel speaking to an SMSC with HTTP and now you want to place an HTTP proxy is in the middle?? Please view in a fixed-width font such as Monaco or Courier. +--+ ++ +-+ | | || | | | KANNEL| | PROXY | | SMSC HTTP | | +-++-+ | | | || | | | | || | | | | || | | +--+ ++ +-+ I would say as long the proxy allows HTTP request to be passed (and of course the network functions properly) and I am assuming that the proxy does not change the HTTP requests. Kannel can reach the SMSC HTTP where it normally should have invoked its HTTP requests on the SMSC directly. Is this what you mean?? Harrie --On Tuesday, September 3, 2002 4:43 AM -0700 Atef lotfy [EMAIL PROTECTED] wrote: Dear Sirs, I have kannel development ver 1.1.6. I want to configure kannel SMSC to be an HTTP,however,Kannel lies behind a proxy. Are the HTTP SMSC http requests forwarded also from the proxy server configurations in the bearer box Or these configurations are only for http requests for SMS service? Please Advice, Atef Lotfy Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server. __ Do You Yahoo!? Yahoo! Finance - Get real-time stock quotes http://finance.yahoo.com Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: [PATCH] SMPP
HI Alan, --On Wednesday, August 14, 2002 10:13 AM +1200 Alan McNatty [EMAIL PROTECTED] wrote: Hi Nisan, Thanks for your patch this is also an issue for me. Can you update docs also - since new config parameter. Does this patch fix also for you the problem?? (You state you have the problem, not that this fixes it) Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: [PATCH] smsc modules directory
--On Sunday, August 11, 2002 12:20 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: Hi Harrie and everyone I saw that smsc.c was moved to the smsc module directory (which I don't think is write - as this is just the API implementation, not a module in itself), but the smsc.h file was left in the gw directory. you have to decide were to put these two files, but they should be together Maybe you are correct. I was in doubt to move the smsc.c file and smsc.h. I did move the smsc.c, since it was used under the smscconn API. Then smsc_wrapper is also moved and I consider that one as an API and a module. It is an API for the old smsc module, but it is a module since it is used from within the smscconn API. I believe this is also due to some legacy and if wanted the wrapper could be complete vapourized if the old smsc module would be ported to the smscconn API. One can argue many ways, but that is also due to the many ways various layes and APIs are made. The include file I left up there, actually incorrect. This counts even so for the smsc_p.h include file. I move them under the smsc directory. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: [PATCH] SMPP panic when sender or receiver address longer then20 characters.
Hi Oded, Good catch. Forgive if I sound stupid, but regarding the length of an address. If it is really bigger then 20 can one just discard the part of the address that is longer?? That sounds weird to me, or is the address only used internal in the smpp part as some indication?? On top of your patch, would it not be better to have the length of an address in some #define and use that?? Then changing the length is done in one place only. --On Sunday, August 11, 2002 1:00 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: Hi list We encountered a bug that causes SMPP to panic when creating the PDU if the receiver or sender address is longer then 20 characters. the panic occures in smpp_pdu.c in smpp_pdu_pack which asserts on the length of the source Octstr before packing it into the PDU. I don't want to change the assert, just make sure that it will not be invoked just because the user enetered bad data, so attached is this patch. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
RE: [PATCH] SMPP panic when sender or receiver address longerthen20 characters.
--On Sunday, August 11, 2002 2:46 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: -Original Message- From: Oded Arbel in some #define and use that?? Then changing the length is done in one place only. it is - in the smpp_pdu.def : hm. my bad - this is not what you meant. doing as you suggested is more correct, but currently that number is only used in two lines of code (one right after the other) and I can't see how we can extract if from the pdu.def, so using it in a define has the single advantage of moving the defining point from halfway down the file to the top of the file (or to the header). I admit these .def files are difficult, but I believe we need to make a define and use that define +1 in the .def file to allcate the space and use the define itself in the other places. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
RE: [PATCH] SMPP panic when sender or receiver address longerthen20 characters.
--On Sunday, August 11, 2002 2:32 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: -Original Message- From: Harrie Hazewinkel [mailto:[EMAIL PROTECTED]] Hi Oded, Good catch. Forgive if I sound stupid, but regarding the length of an address. If it is really bigger then 20 can one just discard the part of the address that is longer?? I don't know - I don't think that there are addresses longer then 20 chars, and if there are - these can not be used with SMPP which only supports addresses as big as 20 chars. I'm not sure of the correct implementation here as I have yet to see any usage examples, but we can hold a hypothetical discussion in the list about how an address longer then 20 chars will look like and how it should be handled if the SMSC does not support such addresses. Not that I want to have a hypothecial discussion, but just dropping parts of an address looks weird to me. Maybe we also have to check were the address is received if it is the correct length. But anyway, making sure it is the correct length like in your patch is a good thing. It also protects against changes within Kannel of the address. That sounds weird to me, or is the address only used internal in the smpp part as some indication?? The receiver address is used as the target MSISDN for the SM and the sender address is used to inform the receiving party of who sent the SM and where to reply too. at least the later can be alpha numeric. On top of your patch, would it not be better to have the length of an address in some #define and use that?? Then changing the length is done in one place only. it is - in the smpp_pdu.def : snip PDU(submit_sm, 0x0004, HEADER NULTERMINATED(service_type, 6) INTEGER(source_addr_ton, 1) INTEGER(source_addr_npi, 1) NULTERMINATED(source_addr, 21) INTEGER(dest_addr_ton, 1) INTEGER(dest_addr_npi, 1) NULTERMINATED(destination_addr, 21) /snip notice the dest_addr and source_addr which are defined as 21 chars null terminated buffers. -- Oded Arbel m-Wise mobile solutions [EMAIL PROTECTED] +972-9-9581711 (116) +972-67-340014 ::.. ...[T]he lesson [comic books] taught children- or this child, at any rate- was perhaps the unintentionally radical truth that exceptionality was the greatest and most heroic of values; that those who were unlike the crowd were to be treasured the most lovingly; and that this exceptionality was a treasure so great that it had to be concealed, in ordinary life, beneath what the comic books called a 'secret identity'. -- Salman Rushdie Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
RE: [PATCH] SMPP panic when sender or receiver address longerthen20characters.
--On Sunday, August 11, 2002 4:14 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: -Original Message- From: Harrie Hazewinkel [mailto:[EMAIL PROTECTED]] I admit these .def files are difficult, but I believe we need to make a define and use that define +1 in the .def file to allcate the space and use the define itself in the other places. K - how about like this ? Cool. I am commiting this. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
CVS and cygwin
HI to all those working and committing from CYGWIN to CVS, I just fixed some file with the '^M' problem. This is caused in most cases by people who use cygwin and CVS. Those need to use a native CVS and not the standard cygwin cvs. The cygwin cvs lets the CVS server think it is a UNIX system, where it is a windows system. more info of native cvs can be found here: http://www.cvshome.org/dev/codewindow.html Hope this helps, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: [PATCH] Octstr function call sometime barf when calledwithNULL data
--On Sunday, August 11, 2002 6:40 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: This will be in my next commit today. I also noticed some other things like ^M' at the end of lines in the ChangeLog and so on. I am going thru all this. this has been commited already. I have to veto against this and would like to request a revision step-back. What is the real problem with it?? I think it is quite a nice solution (used also in other packages I have seen). Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/ Author of MOD-SNMP, enabling SNMP management the Apache HTTP server
Re: [PATCH] SMPP octstr to long handling
--On Sunday, August 11, 2002 6:31 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: BTW, we should not #define constants of SMPP that are explicitely 'defined' by the .def file anyway. So the 21 length for source and destination number is already a #define in this semantical way and hence we should *not* use #defined things in .def again. Hm, I am not so in favour of doing length definitions this hardcoded way. But it seems, that many thing get all the time done via MACROs and depending on the use the marco results in different code. This is more coding style, I guess. RCS file: /home/cvs/gateway/gw/smsc/smpp_pdu.c,v retrieving revision 1.1 diff -r1.1 smpp_pdu.c 136,138c136,147 gw_assert(octstr_len(p-name) max_octets); \ if (p-name != NULL) octstr_append(os, p-name); \ octstr_append_char(os, '\0'); --- if (p-name != NULL) { \ if (octstr_len(p-name) = max_octets) { \ warning(0, SMPP: PDU element %s to long \ (length is %d, should be %d), \ # name, octstr_len(p-name), max_octets); \ temp = octstr_copy(p-name, 0, max_octets-1); \ } else \ temp = octstr_duplicate(p-name); \ octstr_append(os, temp); \ octstr_destroy(temp); \ } \ octstr_append_char(os, '\0'); This also means that you ignore the address part which goes beyond the maximum length. Somehow, this feels not right, or do I miss something?? Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/ Author of MOD-SNMP, enabling SNMP management the Apache HTTP server
Re: [PATCH] Octstr function call sometime barf when called withNULL data
--On Sunday, August 11, 2002 6:38 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: While some Octstr functions handle NULLs gracefully, others (especially the formating ones) do not behave so gentlemen-like. this patch will cause NULLs received instead of Octstr* in some cases not to panic the box but instead be handled in a predictable manner. again -1 for this, sorry Oded. The NULL returns are very dangerious for functions that deal with XML output and hence you dump a null'ed Octstr as NULL cstr which oviously will break XML structures and semantics. I see here the association with XML and I missed that. --On Sunday, August 11, 2002 8:04 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: -Original Message- From: Stipe Tolj [mailto:[EMAIL PROTECTED]] Ok - you are correct here too :-). how about using (NULL) or [NULL] instead ? Hmm, what is exactly the problem with it now (or revision backed) is? I don't got the point I guess. this part : seems_valid_read(ostr,file,line,func); where if the parameter is NULL for some reason, bearerbox panics. I am with Oded here. If a panic can be overcome by the functionality it should be possible. Oded patch did this. Stipe, if you don't want this solution another is to handle an error all the way up to where you can recover, but panicing is to much. And for (NULL) or [NULL] I don't care. Indeed the NULL is to close to XML. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: [PATCH] sms centers
--On Friday, August 9, 2002 12:45 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: so what does the patch do now?! I didn't get the point. It would make the SMSC-code from within the wrapper more equal. Not that it is so important, but afterwards the step to a single modular API is easier. By making everything more equal an API more or less surfaces, since the access to certain parts are then always the same. OK, one could say design a modular API and then enforce it, but that requires the same kind of study on the code/functionality and the change must still be done. This is also why it could be that some patches of me look like 'why is it needed??'. It does not change/fix functionality, but merely move parts of code. That is correct, but it has a goal of ending up with a more modular package. Whether it is worthwhile purchasing this long-term goal I only hope it is worth my efforts. However, this patch is for all the old wrapper SMSC code which are likely to be dropped. Is it still worthwhile applying... Hope this clarifies it, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: creating an smsc directory
--On Friday, August 9, 2002 12:30 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Aarno Syvänen wrote: What about smpp_pdu.* modules ? These are part of smpp implementation. agreed, they should also go to gw/smsc. And don't forget about gw/emimsg.c which is for emi2. Seem to have missed some. I move those too. Already testing the move of those files. Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/ Author of MOD-SNMP, enabling SNMP management the Apache HTTP server
Re: [FYI] critical bug in EMI2 module
--On Saturday, August 3, 2002 6:06 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Hi all, we definitely have a critical bug in the smsc_emi2 module (as far as I have located). When a EMI2 connection gets in a state I can't define currently and the HTTP admin status page is triggered, hence the SMSC list is accessed, the whole bearerbox does a segmentation fault and dumps core. :(( It does not happens as long bearerbox runs and noone tries to access the status page. I'm on the hunt for this one, because it's causing us here at Wapme some very urgent problems with productive SMSC links. How is the locking of the information maintained. A wild guess is that the threads maintaining the connection and associated data is changed during the creation of the status page. In other words some pointer gets changed or some memory block gets freed and the pointer is not made NULL and testing the pointer for not NULL would make you think it is still there. Thus more locking is needed for simultanious access of the data. I noticed that this is a problem while making an SNMP extension. I'd like to release a new development version after we have fixed at least this and maybe 2 or 3 other things. Especially concerning the new HTTP request queueing in smsbox. I'll also try to add a status access information about smsbox's internal state. WOuld do you want to do?? 1) branch of from 1.2.0 release and add the fixes there or 2) take the current head and fix that?? So who's +1 for releasing 1.1.7? Votes from the contributors please. BTW, is the version number plan right with 1.1.7 as development version, which is *newer* than 1.2.0 but the version number is less?! I would advise against this. The numbering scheme is not inconsistent. Almost every one would say 1.2.0 is a newer release then 1.1.7. And this also depends on how and where you want to branch/tag from. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Telefoon simulators for UNIX/Solaris/MacOSX
HI, I was wondering if someone else knows if there are phone simulators that can be used in conjunction with Kannel. I have found various Windows simulators, but I need Unix/Solaris/MacOSX versions. (I don't have a windows system). cheers, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
module API for Kannel
Hi, I have updated the module API I proposed some time ago. I have added thse to the website and you can find the information of it at http://www.kannel.org/module_api/ The txt file is a general design explaination the patch file is a patch that works against a cvs checkout of the date -D2002-8-2 12:00pm Comments good or bad or welcome. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: request for comments: the -Wconversion warnings.
--On Wednesday, July 31, 2002 10:20 PM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Harrie Hazewinkel wrote: The warnings generated by -Wconversion are enormous in amount. I have started to fix some (not even 25%) and one can look in the attached file. What do others think of this?? hmm, what's the benefit from switching from -Wall to all various -Wfoobar?! Most of these are just warnings of declarations which are not correct, duplicate or missing. However, the conversion warnings are a bit overdone maybe, but I know of compilers which gives these warnings default. IMHO, it is wise to check the sizes and so to make them unsigned. However, I guess an unsigned int is enough where in most cases an unsigned long is used. For instance, I also saw once that an octet string function is checked where the length value given as parameter is 0. If so the function returns otherwise does things. However, it it never tested if the value is negative and that was very well possible. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
RE: sms gateway for ucp written is java
--On Thursday, August 1, 2002 2:59 PM +0300 Kaido Karner [EMAIL PROTECTED] wrote: forget about Java if you need a high-performance SMS gateway and take Kannel, contribute to development if it does not fit your needs in a specific way :) that you say .. I remember posting 2 obvious bugfixes .. none of them appear in cvs .. Posted fixes are not always commited in CVS. One of the commiters needs to have the time and understand what and how that perticular big is fixed. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: compiler warnings
HI, --On Wednesday, July 31, 2002 12:53 AM +0100 Bruno David Rodrigues [EMAIL PROTECTED] wrote: On Ter, 2002-07-30 at 10:07, Harrie Hazewinkel wrote: HI, I noticed that Stipe fixed compiler warnings after my patch. So I thought using '--enable-warnings' to have '-Wall' as compile option, but I would like to add additional warnings like '-Wmissing-prototypes' and '-Wmissing-declarations'. Yes, please, add every -W* required to full debug the compile process. I've added --enable-warnings to be able to use the -Wall everytime I make make While adding the various -W* warnings I stumble accros various warnings. I am trying to fix those. But the main one is a signed/unsigned conversion. I believe we better fix these. So, for instance, length and size variable should become all unsigned I could take either unsigned long or size_t and would prefer to use size_t for both is this OK with you all?? Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
request for comments: the -Wconversion warnings.
HI, The warnings generated by -Wconversion are enormous in amount. I have started to fix some (not even 25%) and one can look in the attached file. What do others think of this?? Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server. sus.patch Description: Binary data
compiler warnings
HI, I noticed that Stipe fixed compiler warnings after my patch. So I thought using '--enable-warnings' to have '-Wall' as compile option, but I would like to add additional warnings like '-Wmissing-prototypes' and '-Wmissing-declarations'. Or is the preference towards an extra '--enable-developer' that includes these compile time options?? If I don't hear compliants I add the to the '--enable-warnings'. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: another emi2 bugfix
--On Monday, July 22, 2002 1:34 PM +0300 Kaido Karner [EMAIL PROTECTED] wrote: re emi2 smsc connections by mistake increase received messages counter twice. patch is together with the previous fix in smsc_emi2.c diff -u gateway-1.2.0/gw/smsc_emi2.c gateway-patched/gw/smsc_emi2.c --- gateway-1.2.0/gw/smsc_emi2.cFri Jun 28 14:09:54 2002 +++ gateway-patched/gw/smsc_emi2.c Mon Jul 22 13:29:15 2002 [snip] @@ -1606,8 +1605,6 @@ privdata-deny_ip = deny_ip; if (privdata-rport 0 emi2_open_listening_socket(privdata) 0) { - gw_free(privdata); - privdata = NULL; I think I would prefer to have these combined into a single function. gw_free now maps onto the gw_native_free or gw_check_free with macros. We could make the macros as such that they pass the pointer to the address space containing the pointer and make the pointer even NULL. Then the define #define gw_free(ptr) (gw_native_free(ptr)) would be become #define gw_free(ptr) (gw_native_free(ptr)) and then void gw_native_free(void *ptr); would be become void gw_native_free(void **ptr); What do others think?? I guess that in more places the pointer needs to be 'NULL'-ed after the free. Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/ Author of MOD-SNMP, enabling SNMP management the Apache HTTP server
RE: another emi2 bugfix
--On Tuesday, July 30, 2002 12:30 PM +0300 Kaido Karner [EMAIL PROTECTED] wrote: I think I would prefer to have these combined into a single function. gw_free now maps onto the gw_native_free or gw_check_free with macros. We could make the macros as such that they pass the pointer to the address space containing the pointer and make the pointer even NULL. What do others think?? I guess that in more places the pointer needs to be 'NULL'-ed after the free. I'd say no - this would propagate bad programming practices. As for bad practises, you can see it that way. I however would say it maps onto another function which can do more the 'free' alone. IMHO, that warrants the change. It also protects developers. What do others think?? Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
logging in Kannel
HI, In order to make things more modular I have looked into the way logging is handled in Kannel. To my horrible discovery there are two kinds of logging in gwlib namely, access_log.[c|h] and log.[c|h]. Accesslog is supposed to be a more simple API/function as the log version. I also noticed accesslog is not even used in the wap portion for instance. So first question is: Why is only in smsbox and bearerbox the accesslog used and not in the wap portion?? No one interested in the accesslog there?? Or are people trying to get the access log part out of a generic logfile containing also many (potentially) debug logs. second question: Would it not be easier to provide to the alog function only the Msg structure (and maybe an extra note) where the alog function determines the output format?? Now a 'const chat*fmt,' is given. OK, this looses flexibility, but it can unify all the formats going into the accesslog. But a more common API would allow for multiple logpaths when this layered. Meaning firs the common logfunction is called in there depending on the module a different logfunction. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: A small contribution to Kannel 1.1.6
--On Friday, July 19, 2002 1:22 PM +0100 James Wang [EMAIL PROTECTED] wrote: I proposed the above change on May 9, 2002. However, I found that the latest release, Kannel 1.2.0, still have not included this patch. Can anyone explain please? I did not look at the originally porposed change, but if you would propose a patch that can be applied instead of a description on what to do that helps. (making a patch diff -ru orig new patchfile or something that applies to your environment) I believe many are busy with many things, so if you can make it as easy for those who have commit access (by providing a patch) it maybe will be picked up. Hope this helps, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
after the release
HI all, Since we now have a new release 1.2.0 I assume we will develop on head again and do bugfixes in a branch. I laid low for a while waiting for the release, but I am now going to add various changes again. My changes are in perticular aimed in making it all more modular. Which sometimes just means making things more equal and then combine those. I will use the STATUS file to keep you up to date. Major design changes I first will distribute to the list. Harrie --- Author of MOD-SNMP, enabling SNMP management to the Apache server.
Re: after the release
HI, --On Monday, July 8, 2002 8:06 PM +0200 Harrie Hazewinkel [EMAIL PROTECTED] wrote: I will use the STATUS file to keep you up to date. Major design changes I first will distribute to the list. Sorry, I ment Changelog. Harrie
Re: smsbox terminating
HI, Which version of Kannel is this?? 1.1.6, 1.2.0rc1 or CVS?? --On Wednesday, July 3, 2002 8:33 PM +0500 Cipher Strength [EMAIL PROTECTED] wrote: Operating System Redhat: 7.3 Kannel Development Release Following is the smsbox last log 2002-06-25 15:16:13 [0] INFO: Connection closed by the bearerbox 2002-06-25 15:16:13 [0] INFO: Received (and handled?) 0 requests in 928709 secon ds (0.00 per second) 2002-06-25 15:16:13 [0] INFO: Kannel smsbox terminating. 2002-06-25 15:16:13 [1] WARNING: Destroying fdset with 996 active entries. CIPHER _ Send and receive Hotmail on your mobile device: http://mobile.msn.com Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
Re: Kannel's module API
(resend, since it bounced due to wrong sender address) Hi Stipe, --On Tuesday, June 11, 2002 12:24 AM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Hi Harrie, I liked the document you wrote about Kannel's API structure. Hope that more of the other developers have a chance to review it. I'll have a deaper look into the patch itself. The patch itself needs some updating, since some other changes cause collisions. I will update it in a day or 2. IMO, the way apache-1.3 does the modularization may be addopted to Kannel for those API purposes. What do the the others think of this? It is indeed based on that :-)) We even could do it based on the Apache 2.0 api, but that is somehow trivial. The concept for now is more important. Should we start re-constructing the current devel tree after we released 1.2.0 and keep a stable branch? I'm +1 for re-constructing Not sure, what you mean here. But I assume you want to branch for the 1.2.0 releases and development goes into head, right?? a) to a general autoconf layout, hence utilize libtool (yet no assigned volonteers here) Using libtool is not directly my goal, but it seems to make things as porting easier. I also believe that making the build process a bit more fixed is not a bad thing. Now it creates some libraries of arbitrary files and things can easily extended where a new 'gmake .depend' seem to add b) for modularization using a Kannel API structure (Harrie will be leading this effort, because he's highly involved in Apache's API structure and has most experiences here IMO) I don't have Kannel as some kind of day-job, unlike some of the others here. Meaning as soon I have some paid-work that will get more priority. But I am willing to lead this effort. I even would not mind working on this for some company who needs it. :-)) BTW, I'm working on a new Kannel web site (using PHP4 and MySQL) and other funny things ;) Cool. What about ssh on the cvs server?? That would make things more secure on that part too. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
Re: Daily patch: gateway
--On Friday, May 31, 2002 10:17 AM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Angel, File gateway/gw/bb.h changed from revision 1.5 to 1.6 File gateway/gw/bb_smscconn.c changed from revision 1.45 to 1.46 File gateway/gw/heartbeat.c changed from revision 1.2 to 1.3 File gateway/gw/heartbeat.h changed from revision 1.1 to 1.2 File gateway/gw/smsbox.c changed from revision 1.198 to 1.199 File gateway/gw/smsc_smpp.c changed from revision 1.67 to 1.69 File gateway/gw/wapbox.c changed from revision 1.148 to 1.149 All this files have been changed without documentation on ChangeLog. Some of those, I chnaged and I did not edit ChangeLog, since I thought that was updated automatically with the CVS message provided at commit. I will put the same message in ChangeLog. Sorry about that. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
RE: Daily patch: gateway
--On Friday, May 31, 2002 12:07 PM +0200 Angel Fradejas [EMAIL PROTECTED] wrote: Harrie please, Would you be so kind as to comment these changes in ChangeLog? Done. They're somewhat described in a post from May 25 (subject: changes to the heartbeat code.) The message is the same as the commit message. I personally, would prefer if the CVS-message during commit is used for this. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
Re: question regarding the configuration implementation
--On Friday, May 31, 2002 9:38 AM +0200 Stipe Tolj [EMAIL PROTECTED] wrote: Harrie Hazewinkel wrote: I was wondering if someone could tell me the reason why putting it in a dictionary first as a very general configuration representation and later search into it for only those parts needed?? I'm not sure, but is the dictionary needed for faster lookups later when the groups are needed. OK, but what is fast, you only do this at startup. :-) The thing I merely was wondering of is why a 2 step phase. First loading it all in the process and that as a second phase look into the values of it. See also below. Another question I have is why one would have all configuration tokens known to each process and depending on the process totally ignore it?? The main reason is maybe that this way a single configuration file can be used for all, but I guess there is more. Hmm, I didn't understand the question Harrie. I will explain it again. The cfg.def, cfg.c and cfg.h file in the gwlib are the core of the configuration handling. All tokens are known there. Those who are not will cause an error while reading the configuration file. All others are put into a list/dictionary for later use. The later use by the process is just to call the ones it needs from the list/dictionary. All others are basically ignored. One could make it like this. Each process only knows those tokens it will use. All other tokens will generate an error. This as oppose to read all possible tokens of any of the processes and just ignore those not used by that process. Therefore, my question 'why should each process know all possible tokens for all processes of the package and those not needed being ignored by the process'?? I believe this is done so only one configuration file is needed for the complete system (bearebox and wap/smsbox). Howevr, this would make a more module system problematic, since not all processes need the same module and thus do not know all the same tokens. Hope this explains it better. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
RE: CVS usage question??
--On Wednesday, May 29, 2002 5:54 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: It's possible, if your editing tool reformats tabs and removes spare spaces and such. try to diff all the changes agains the current cvs, and then edit the diffs and remove everything that isn't requires for the patch. then you can checkout from the CVS again, apply the diffs you've edited and commit. The problem is ( I believe) that I do not see any difference with a cvs diff. Except for those files I edited, but they show up all on the list with cvs commit. Could it be a problem that I have seen files with a ^M that are somehow changes by cvs?? Alternatively, one can do a 'cvs up' to update and just see which files are modified. That all looks OK to me. Alternativly you can either switch to another code editor, or - commit all the files anyway and hope that your editor will not change any more files. I use 'vi'. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
RE: CVS usage question??
--On Wednesday, May 29, 2002 5:41 PM +0200 Angel Fradejas [EMAIL PROTECTED] wrote: Maybe line-feed format has been changed from unix-like to dos-like. This could happen if you use Windows based tools to edit, etc. that could be it, see also my other email in which I mentioned the '^M'. However, I noticed also that many files have a '^M' at the end. So what would we prefer? Angel. -Mensaje original- De: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]En nombre de Oded Arbel Enviado el: miércoles 29 de mayo de 2002 16:54 Para: Harrie Hazewinkel; dev-kannel Asunto: RE: CVS usage question?? It's possible, if your editing tool reformats tabs and removes spare spaces and such. try to diff all the changes agains the current cvs, and then edit the diffs and remove everything that isn't requires for the patch. then you can checkout from the CVS again, apply the diffs you've edited and commit. Alternativly you can either switch to another code editor, or - commit all the files anyway and hope that your editor will not change any more files. -- Oded Arbel m-Wise Inc. [EMAIL PROTECTED] (972)-67-340014 (972)-9-9581711 (ext: 116) ::.. Was it Ritchie or Thompson who said about X: Sometimes, when you fill a vacuum, it still sucks? -Original Message- From: Harrie Hazewinkel [mailto:[EMAIL PROTECTED]] Sent: Wednesday, May 29, 2002 2:29 PM To: dev-kannel Subject: CVS usage question?? HI all, I wanted to commit changes in the gw directory. Although, only 5 of the many files were changed. But in the 'cvs commit' it showed almost all files to be chnaged. Has anyone seen it before?? Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/
RE: Patch submission and release policy (Was: [PATCH] problemswithHTTPS and base support for per message billing)
--On Wednesday, May 29, 2002 4:41 PM +0300 Oded Arbel [EMAIL PROTECTED] wrote: Maybe it is also needed to establish of how many of the group should support or oppose to it before it is decided to what will be added. Something like this is needed when you have a one branch development. when you branch the developers away from the stable release - why bother ? everyone who has CVS access can commit features as long as care is take not to break anything and proper notification is done (current devel-reports is more then enought - just remember to document your changes in ChangeLog). only conceptual/architectual changes need to be discussed in the open. Changelog is generated from the cvs commit messages or is it neccessary to do it by hand?? Just wondering, since I thought ChangeLog can be generated from CVS itself. (I could be wrong) - branch the tree now (yesterday would have been a good time too ;-) and label it 1.2.0. Why do you already want a branch?? Why not putting the releases into a branch?? (Just asking) Since this way you just can keep developing. As for a release I would suggest name it 1.2.0 Make a branch of it and do new releases as 1.2.x Continue development in the main branch for 1.3.y. Semantics. I don't mind what you call it, as long as we have a stable branch and a development branch. Fine, but I would think that you would at least want a different branch for for instance the 1.2.x and the 1.3.y product/stable release. But for that, before thinking in branches and releases, we should think in the new architecture. I concur. The only thing is what kind of things do we put on the list. You already mentioned some. After what we have to define how indeed. Why bother - you can't make architectual changes w/o branching the tree - you need to get a declared stable release before rewriting the entire code base, so people will have something to put on their production servers while the project's next generation is in limbo. What people should put on their production systems is something tested and released as stable. I don't think that people should just take a CVS checkout for the production system. Those poeple take a risk where a released version should have been tested before release. IMHO, that has not much to do with branching. Besdies that I intended to say here that we first need to discuss how the architecture will be and some of those other things. After that we need to decide on the implementation and how to use branches. Maybe some things are having very little impact and thus do not need a special (own) branch. Harrie Internet Management Consulting mailto: [EMAIL PROTECTED] http://www.lisanza.net/
RE: version 1.1.6 processing limits
--On Thursday, May 30, 2002 10:14 AM +0300 Oded Arbel [EMAIL PROTECTED] wrote: I'm not sure how native malloc works, but the checking malloc (which we use always) has an upper limit on the number of allocations allowed. if you want to use more memory you will have to recompile. Just a thought, but is the amount of system-call casuing some problems. Not knowing how much there are, but malloc is one of them. If so, maybe some memory handling needs to be added which does once allocates a huge block from the system and then in user land claims pieces where needed. At the end of some cycle all is returned at once, instead of all those small bits. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
changes to the heartbeat code.
Hi, I have attached a patch to group the heartbeat code. If I hear/see no objections I will commit this. It changes: 1) adds a stop all heartbeat functionality by by means of the heartbeat_thread value of '-1' (ALL_HEARTBEATS) as the heartbeat_stop function. 2) move the heartbeat frequentie default value in the heartbeat.h (IMHO, where it belongs) 3) changes the wapbox.c and smsbox.c where the returned thread_nr is only used for providing a warning and the heartbeatthread_stop does a stop ALL_HEARTBEATS. regards, Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ heartbeat.patch Description: Binary data
CVS usage question??
HI all, I wanted to commit changes in the gw directory. Although, only 5 of the many files were changed. But in the 'cvs commit' it showed almost all files to be chnaged. Has anyone seen it before?? Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
RE:first draft and patch for module oriented API
--On Wednesday, May 29, 2002 12:18 PM +1000 Ignat Vassilev [EMAIL PROTECTED] wrote: Hi Harrie I tried to compile kannel with modules.patch but i recive error Hmm, I will try top built a patch again. I tested it here a few times and compilation was OK. Except for the test tools. I did not bother to mamke changes there. 1) Did you do a 'rm .depend ; gmake .depend'?? Otherwise, the new files (resulting in an object-file) do not end up in the libraries. 2) It also could be that the required patch lines (that includes the file bearerbox_modules.h) in the bearer_box.c did not went correctly. 3) It was a patch for the 'HEAD of CVS'. Hang on, I will see what really goes wrong. Since I did a change in CVS, the patch will change as well. New one will be provided soon. gcc -D_REENTRANT=1 -I. -g -O2 -DBROKEN_PTHREADS=1 -I/usr/include/libxml2/libxml -I/usr/include/libxml2 -I/include -Wall -I/usr/include/openssl -I/usr/include/mysql -o gw/bearerbox gw/bearerbox.o libgw.a libwmlscript.a libwap.a libgwlib.a -lmysqlclient -lssl -ldl -lpam -lpthread -lresolv -lnsl -lm -L/usr/lib -lxml2 -lz -L/lib -lm -L/usr/lib -lcrypto -lssl -L/usr/lib/mysql -lmysqlclient libgwlib.a(modules.o): In function `is_allowed_by_module': /home/ozzy/kannel/gateway/gwlib/modules.c:54: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:56: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:57: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:67: undefined reference to `module_array_size' libgwlib.a(modules.o): In function `run_init': /home/ozzy/kannel/gateway/gwlib/modules.c:84: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:86: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:87: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:93: undefined reference to `module_array_size' libgwlib.a(modules.o): In function `run_start_thread': /home/ozzy/kannel/gateway/gwlib/modules.c:104: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:107: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:109: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:110: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:116: undefined reference to `module_array_size' libgwlib.a(modules.o): In function `run_stop_thread': /home/ozzy/kannel/gateway/gwlib/modules.c:126: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:129: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:131: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:132: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:134: undefined reference to `module_array_size' libgwlib.a(modules.o): In function `run_log': /home/ozzy/kannel/gateway/gwlib/modules.c:145: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:147: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:148: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:154: undefined reference to `module_array_size' libgwlib.a(modules.o): In function `run_exit': /home/ozzy/kannel/gateway/gwlib/modules.c:164: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:166: undefined reference to `module_array_size' /home/ozzy/kannel/gateway/gwlib/modules.c:167: undefined reference to `module_array' /home/ozzy/kannel/gateway/gwlib/modules.c:169: undefined reference to `module_array_size' collect2: ld returned 1 exit status make: *** [gw/bearerbox] Error 1 Regards Ignat Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
RE: Patch submission and release policy (Was: [PATCH] problems withHTTPS and base support for per message billing)
--On Wednesday, May 29, 2002 11:00 AM +0300 Oded Arbel [EMAIL PROTECTED] wrote: Well.. To be honest, using the CVS is an advantage because that way we get 100% testing and debug, code is done with less errors and bugs are fixed quicker ;) I'm always using cvs in production. Some bugs are only visible on production systems and I don't have time to do testings before upgrading. And if some message is lost, I can always blame the SMSC ;) There's some structural changes that we should do, and for that we really need a different branch. Modularity, new autoconf, real unicode support, etc. But for that, before thinking in branches and releases, we should think in the new architecture. I do not agree - we cannot even think of architecture changes while everyone is building their production systems from CVS. I agree here. while here we also build our production from CVS by choice, for the same reasons you stated, this is a bad thing(tm). it's our obligation to supply a stable branch for people who rather have something that is known to work, then the bleeding edge (which is most people). I would say (as mentioned in other emails) take the branch for a stable release version. Those should not change that much as the main branch were development is done. If we start doing architectural changes on the CVS, while everyone is using it to build their production, we will break things for people who don't/can't know how to handle it. I simply do not understnad why people in the first place use a, so-called, non stable CVS version in the production. It is well possible that it si not even tested well enough. So a branch is a must before doing any major surgery on the code. I agree. However, I believe that the main branch should be the main development line, unless it would be something very experimental. Harrie Internet Management Consulting mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
first draft and patch for module oriented API.
HI All, Attached is a document slightly explaining a module oriented API and a patch assoicated for it. I would people like to invite to make comments and in perticular those who made additional modules that need always patching the core. I understand it is not complete yet, but that is why I invite comments. Harrie Internet Management Consulting tel: +39-3474932300 mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ modules.patch Description: Binary data module_struct.txt Description: Binary data
question regarding the configuration implementation
HI all, Maybe someone could help me out in order to confirm how I think the configuration file is hanndled. I believe this is the scheme: A programm 1- opens the configuration file 2- reads it as a huge octet_string 3- splits is into a list of lines (octet_string) 4- parser the all lines of the list which could be a- a comment line (ignore) b- an include line (which inserts the to be included as a list of lines again steps 1,2,3) c- processes a lines (splits is at the '=', makes a name value pair and adds it in a dictionary) (NOW ALL CONFFILE READING IS DONE) 5- initializes the needed process a- it starts for looking of groups in the dictionary b- depending on the group it looks for parameters. I was wondering if someone could tell me the reason why putting it in a dictionary first as a very general configuration representation and later search into it for only those parts needed?? Another question I have is why one would have all configuration tokens known to each process and depending on the process totally ignore it?? The main reason is maybe that this way a single configuration file can be used for all, but I guess there is more. I merely ask this on how to handle module specific configuration parts for that module API I was talking of last week. Any help is appreciated, Harrie Internet Management Consulting tel: +39-3474932300 mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
[PATCH] typo spesified
HI, A typo some in the files, spesified = specified. regards Harrie Internet Management Consulting tel: +39-3474932300 / +31-625357135 mailto: [EMAIL PROTECTED] http://www.lisanza.net/ specified.patch Description: Binary data
idea of multi purpose logging
HI, I was wondering what people would think of the idea of a multi purpose logging. Currently (in version 1.1.6) one can only use logfiles in a syslog style or log to stdout/stderr. So what do I mean by multi purpose logging: A wrapper of all kinds of logging possible from where to a specific type of logging can be done. This would allow in a more clean way more kinds of logging in such a way that the core code still only has only call and from within that call the multiple logging functions are done. As a result the gwlib/log.c is the only file that needs to be changed and there needs to be a log-function file for each log function. I hope I have made myself clear if not say so and I will elaborate a bit more. regards, Harrie Internet Management Consulting tel: +39-3474932300 / +31-625357135 mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
some other warnings of Kannel
HI, (repost due to Non-members are not allowed to post messages to this list (to avoid spam) I have compiled the Kannel version 1.1.6 on FreeBSD 4.1-RELEASE i386 and got these warnings. Read the last one :-)) cc -funsigned-bitfields -D_REENTRANT=1 -I. -g -O2 -I/usr/local/include/libxml2 -o checks/check_octstr checks/check_octstr.o libgw.a libwmlscript.a libwap.a libgwlib.a -lm -lc_r -L/usr/local/lib -lxml2 -lz -lm /usr/lib/libc.so: WARNING! setkey(3) not present in the system! /usr/lib/libc.so: warning: this program uses gets(), which is unsafe. /usr/lib/libc.so: warning: mktemp() possibly used unsafely; consider using mkstemp() /usr/lib/libc.so: WARNING! des_setkey(3) not present in the system! /usr/lib/libc.so: WARNING! encrypt(3) not present in the system! /usr/lib/libc.so: warning: tmpnam() possibly used unsafely; consider using mkstemp() /usr/lib/libc.so: WARNING! des_cipher(3) not present in the system! /usr/lib/libc.so: warning: tempnam() possibly used unsafely; consider using mkstemp() /usr/lib/libc.so: warning: this program uses f_prealloc(), which is stupid. Harrie Internet Management Consulting tel: +39-3474932300 / +31-625357135 mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/
[PATCH] counter type change
Hi, (repost due to using a non-subscribed address) I have attached a patch for the kannel code which make the internal type of the counter object an 'unsigned long'. The advantage is that it wraps by itself and we have 1 bit more in the value of the counter (32 bits). :-)) I noticed also that sometimes the value of the counter returned by the accessor-functions used already 'unsigned long'. I have also added a 'counter_increase_with' value for usage to increase counters by more then 1. This is when one would count for instance the total bytes of the messages sent. Besdies I have a question; why is in most cases the counter value retrieved and stored elsewhere. One could directly get the value of the counter by just keeping the reference to it in those placed where now this value is 'copied'. cheers, Harrie Internet Management Consulting tel: +39-3474932300 / +31-625357135 mailto:[EMAIL PROTECTED]http ://www.mod-snmp.com/ counter.patch Description: Binary data