Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Jan 22 12:46, Jason Tishler wrote: > I am *very* happy to report that PostgreSQL running against cygserver > passes all regression tests under the 2004-Jan-21 snapshot. Good news. Thanks for the report, Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Developermailto:[EMAIL PROTECTED] Red Hat, Inc.
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Corinna, On Sat, Jan 03, 2004 at 10:08:26AM -0500, Jason Tishler wrote: > I have some good news to report. First, PostgreSQL builds OOTB > against cygserver without *any* source changes. Second, PostgreSQL > appears to work just fine when used casually. > > Unfortunately, when I run PostgreSQL's regression test (which drives > the system "hard"), I get the following errors: > > [snip] > > Unfortunately, my build of the latest Cygwin CVS does not appear to > work -- all commands just immediately exit. I will try to debug > further when I can build a working version of the DLL. I am *very* happy to report that PostgreSQL running against cygserver passes all regression tests under the 2004-Jan-21 snapshot. I really appreciate your efforts to complete the Sys V IPC support work. Thanks, Jason -- PGP/GPG Key: http://www.tishler.net/jason/pubkey.asc or key servers Fingerprint: 7A73 1405 7F2B E669 C19D 8784 1AFD E4CC ECF4 8EF6
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Fri, Jan 09, 2004 at 04:52:39PM +0100, Corinna Vinschen wrote: >Hmm, my cygserver is cross-build and I don't see this sort of problem. >My build is from 2003-12-16. Huh? You know that we are using different compilers. We've traded several private messages about this. You *know* that I can't use a cygserver that I've built on linux. cgf
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Jan 9 10:49, Christopher Faylor wrote: > On Fri, Jan 09, 2004 at 03:50:41PM +0100, Corinna Vinschen wrote: > >On Jan 9 08:36, Jason Tishler wrote: > >> Corinna, > >> > >> On Wed, Jan 07, 2004 at 11:46:18AM +0100, Corinna Vinschen wrote: > >> > On Jan 3 10:08, Jason Tishler wrote: > >> > > I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest > >> > > CVS as my test environment. > >> > > >> > Would you mind to try with the snapshot from 12-18 just for the sake > >> > of testing without the latest signal changes? > >> > >> I'm sorry to say that the 12-18 snapshot is even worse. :,( > > > >Can you try to find out why cygserver just exits? A first idea is to > >switch debugging on: > > > >$ /usr/sbin/cygserver --debug > > It just occurred to me that this problem is probably due to the > mysterious failures that I see with cross-compiled cygservers. If you > build cygserver natively it seems to work better. Hmm, my cygserver is cross-build and I don't see this sort of problem. My build is from 2003-12-16. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Developermailto:[EMAIL PROTECTED] Red Hat, Inc.
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Fri, Jan 09, 2004 at 03:50:41PM +0100, Corinna Vinschen wrote: >On Jan 9 08:36, Jason Tishler wrote: >> Corinna, >> >> On Wed, Jan 07, 2004 at 11:46:18AM +0100, Corinna Vinschen wrote: >> > On Jan 3 10:08, Jason Tishler wrote: >> > > I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest >> > > CVS as my test environment. >> > >> > Would you mind to try with the snapshot from 12-18 just for the sake >> > of testing without the latest signal changes? >> >> I'm sorry to say that the 12-18 snapshot is even worse. :,( > >Can you try to find out why cygserver just exits? A first idea is to >switch debugging on: > >$ /usr/sbin/cygserver --debug It just occurred to me that this problem is probably due to the mysterious failures that I see with cross-compiled cygservers. If you build cygserver natively it seems to work better. So the cygserver.exe in any snapshot is probably crap. I haven't had time to bail this particular part of the boat out recently since I'm bailing another section right now. cgf
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Jan 9 08:36, Jason Tishler wrote: > Corinna, > > On Wed, Jan 07, 2004 at 11:46:18AM +0100, Corinna Vinschen wrote: > > On Jan 3 10:08, Jason Tishler wrote: > > > I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest > > > CVS as my test environment. > > > > Would you mind to try with the snapshot from 12-18 just for the sake > > of testing without the latest signal changes? > > I'm sorry to say that the 12-18 snapshot is even worse. :,( Can you try to find out why cygserver just exits? A first idea is to switch debugging on: $ /usr/sbin/cygserver --debug Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Developermailto:[EMAIL PROTECTED] Red Hat, Inc.
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Corinna, On Wed, Jan 07, 2004 at 11:46:18AM +0100, Corinna Vinschen wrote: > On Jan 3 10:08, Jason Tishler wrote: > > I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest > > CVS as my test environment. > > Would you mind to try with the snapshot from 12-18 just for the sake > of testing without the latest signal changes? I'm sorry to say that the 12-18 snapshot is even worse. :,( Jason -- PGP/GPG Key: http://www.tishler.net/jason/pubkey.asc or key servers Fingerprint: 7A73 1405 7F2B E669 C19D 8784 1AFD E4CC ECF4 8EF6
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Hi Jason, On Jan 3 10:08, Jason Tishler wrote: > Corinna, > > I *very* sorry for the very sluggish response time. It is not a No worries, I was on vacation, together with a flu, anyway. > I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest > CVS as my test environment. Would you mind to try with the snapshot from 12-18 just for the sake of testing without the latest signal changes? Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Developermailto:[EMAIL PROTECTED] Red Hat, Inc.
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Corinna, I *very* sorry for the very sluggish response time. It is not a reflection on your great work or a lack of interest on my part -- it is just the sad fact that my life's hectic factor went up by a magnitude or two over its normal feverish pace for the last month or so... :,( On Wed, Nov 26, 2003 at 12:10:58PM +0100, Corinna Vinschen wrote: > a headsup which affects a few packages and, of course, the developers > of Cygwin. Last week I have applied a pretty big patch, which turns > Cygserver upside down. It now contains basically the FreeBSD kernel > implementation of System V (or, in X/Open / SUSv3 speak) XSI IPC support. > > This means, not too far from now, we can drop the cygipc package and > packages like postgresql can switch to using Cygserver. > > [snip] > > However, I'd like to ask maintainers of affected packages to give > Cygserver a try. You will have to rebuild your packages and you > should make sure that the Cygwin header files sys/ipc.h, sys/msg.h, > sys/sem.h and sys/shm.h overwrote the cygipc files before building > your package(s). And don't link against libcygipc ;-) I have some good news to report. First, PostgreSQL builds OOTB against cygserver without *any* source changes. Second, PostgreSQL appears to work just fine when used casually. Unfortunately, when I run PostgreSQL's regression test (which drives the system "hard"), I get the following errors: 3 [main] postmaster 1144 transport_layer_pipes::connect: lost connection to cygserver, error = 2 LOG: server process (PID 1144) was terminated by signal 12 LOG: terminating any other active server processes LOG: all server processes terminated; reinitializing 371 [main] postmaster 2056 transport_layer_pipes::connect: lost connection to cygserver, error = 2 LOG: semctl(65536, 0, IPC_RMID, ...) failed: Interrupted system call LOG: shmdt(0xbe) failed: Interrupted system call LOG: shmctl(65536, 4096, 0) failed: Interrupted system call FATAL: could not create shared memory segment: Interrupted system call DETAIL: Failed system call was shmget(key=5432001, size=8970240, 03600). I believe the above occurs because cygserver "just" exits: $ /usr/sbin/cygserver cygserver: Initialization complete. Waiting for requests. $ I am using cygwin-inst-20031231.tar.bz2 with cygserver from the latest CVS as my test environment. FWIW, fetchmail seems to "just" exit when run under this version of Cygwin too. So, maybe the problem is with Cygwin and not cygserver? Unfortunately, my build of the latest Cygwin CVS does not appear to work -- all commands just immediately exit. I will try to debug further when I can build a working version of the DLL. Thanks, Jason -- PGP/GPG Key: http://www.tishler.net/jason/pubkey.asc or key servers Fingerprint: 7A73 1405 7F2B E669 C19D 8784 1AFD E4CC ECF4 8EF6
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Wed, Nov 26, 2003 at 10:38:29PM -0500, Harold L Hunt II wrote: > Christopher Faylor wrote: > >The gotcha here, though, is that you still have to install cygserver > >and get it running. Corinna has provided install scripts to make > >that to happen but experience shows us that there will probably be > >some tweaking required before everything is perfect. You just have the same gotcha with cygipc. So nothing really new with this. > Ah, well, in the case of XWin.exe, it detects whether the an SHM > allocation succeeds or fails and enables/disables the X SHM extension > appropriately. Each X application tests for presence of the SHM > extension and doesn't try to use it if it is not there. > > Thus, all of our apps and libs link against cygipc. We can still build > with SHM support but it won't actually be used unless things are setup > properly, as you described. Ok, I don't know how that's implemented in cygipc so I think it might be necessary to mention that: The implementation of the MSG, SEM and SHM functions in Cygwin are so that if the functions are not available (be it that CYGWIN doesn't contain the word "server" or Cygserver isn't running) not only set errno to ENOSYS, but they also raise a SIGSYS. So it's crucial that your code handles that signal. At least, it should ignore it (signal(SISYS, SIG_IGN)). Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Developermailto:[EMAIL PROTECTED] Red Hat, Inc.
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Christopher Faylor wrote: On Wed, Nov 26, 2003 at 07:40:25PM -0500, Harold L Hunt II wrote: So, there is no "cygserver" package... it will just be part of cygwin1.dll? Is that correct? Excuse me for not following development until now... it really hasn't affected me until today. Right. cygserver.exe will be part of the cygwin package. Okay. Thanks. The cygserver directory also contains a README file which describes Where is this cygserver directory? Are you referring to the CVS tree for cygwin1.dll? Yes. Thanks. However, I'd like to ask maintainers of affected packages to give Cygserver a try. You will have to rebuild your packages and you should make sure that the Cygwin header files sys/ipc.h, sys/msg.h, sys/sem.h and sys/shm.h overwrote the cygipc files before building your package(s). And don't link against libcygipc ;-) Sounds like we are going to have to have a flag day for Cygwin/X and all related packages to remove the link to cygipc and use the built-in SHM support provided by cygserver. That is not really a problem for me, and I know more about package maintanence now so it should be easier this time around. The gotcha here, though, is that you still have to install cygserver and get it running. Corinna has provided install scripts to make that to happen but experience shows us that there will probably be some tweaking required before everything is perfect. Ah, well, in the case of XWin.exe, it detects whether the an SHM allocation succeeds or fails and enables/disables the X SHM extension appropriately. Each X application tests for presence of the SHM extension and doesn't try to use it if it is not there. Thus, all of our apps and libs link against cygipc. We can still build with SHM support but it won't actually be used unless things are setup properly, as you described. That is okay: most people don't need the SHM support. The only reason for SHM support in the default server is to keep from having to distribute both an SHM-enabled and SHM-disabled set of packages; now we have one set that can use SHM if it is setup properly. So I think we will be okay... no one will be required to have cygserver actually working. Harold
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
On Wed, Nov 26, 2003 at 07:40:25PM -0500, Harold L Hunt II wrote: >Corinna, > >Corinna Vinschen wrote: >>This means, not too far from now, we can drop the cygipc package and >>packages like postgresql can switch to using Cygserver. > >Additionally, most XFree86-* packages and anything that depends upon >XFree86-bin or XFree86-base (cgoban, emacs-X11, ghostscript-x11, grace, >gv, and tetex-x11). > >>In future, Cygwin releases will be build with Cygserver support switched >>on. Clients, which want to access the Cygserver capabilities, have to >>run with the environment variable CYGWIN set so that it contains the >>word "server". I guess this will be the default at some later point. > >So, there is no "cygserver" package... it will just be part of >cygwin1.dll? Is that correct? Excuse me for not following development >until now... it really hasn't affected me until today. Right. cygserver.exe will be part of the cygwin package. >>The cygserver directory also contains a README file which describes > >Where is this cygserver directory? Are you referring to the CVS tree >for cygwin1.dll? Yes. >>However, I'd like to ask maintainers of affected packages to give >>Cygserver a try. You will have to rebuild your packages and you >>should make sure that the Cygwin header files sys/ipc.h, sys/msg.h, >>sys/sem.h and sys/shm.h overwrote the cygipc files before building >>your package(s). And don't link against libcygipc ;-) > >Sounds like we are going to have to have a flag day for Cygwin/X and all >related packages to remove the link to cygipc and use the built-in SHM >support provided by cygserver. That is not really a problem for me, and >I know more about package maintanence now so it should be easier this >time around. The gotcha here, though, is that you still have to install cygserver and get it running. Corinna has provided install scripts to make that to happen but experience shows us that there will probably be some tweaking required before everything is perfect. cgf
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Corinna, Corinna Vinschen wrote: This means, not too far from now, we can drop the cygipc package and packages like postgresql can switch to using Cygserver. Additionally, most XFree86-* packages and anything that depends upon XFree86-bin or XFree86-base (cgoban, emacs-X11, ghostscript-x11, grace, gv, and tetex-x11). In future, Cygwin releases will be build with Cygserver support switched on. Clients, which want to access the Cygserver capabilities, have to run with the environment variable CYGWIN set so that it contains the word "server". I guess this will be the default at some later point. So, there is no "cygserver" package... it will just be part of cygwin1.dll? Is that correct? Excuse me for not following development until now... it really hasn't affected me until today. The cygserver directory also contains a README file which describes Where is this cygserver directory? Are you referring to the CVS tree for cygwin1.dll? However, I'd like to ask maintainers of affected packages to give Cygserver a try. You will have to rebuild your packages and you should make sure that the Cygwin header files sys/ipc.h, sys/msg.h, sys/sem.h and sys/shm.h overwrote the cygipc files before building your package(s). And don't link against libcygipc ;-) Sounds like we are going to have to have a flag day for Cygwin/X and all related packages to remove the link to cygipc and use the built-in SHM support provided by cygserver. That is not really a problem for me, and I know more about package maintanence now so it should be easier this time around. Harold
Re: HEADSUP: cygserver now has MSG, SEM and SHM support
Corinna Vinschen wrote: > This means, not too far from now, we can drop the cygipc package and > packages like postgresql can switch to using Cygserver. Woo Hoo This just MADE my holiday! Thanks, Corinna!! You ROCK! -- Chuck
