Re: [PATCH]? Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
David Kimdon wrote: >Very annoying one is 123382 (kernel-image-2.2.20-idepci: >need CONFIG_PACKET=y for boot-floppies dhcp-client, which I'll re-open >and reassign to boot-floppies shortly). Hmm, what's the deal here? We already put af_packet.o on the root disk for the idepci, compact and vanilla flavours, and there should be stuff in rcS to load it. Can you try insmod'ing it by hand and see if you can figure out what's up? p. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: [PATCH]? Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
> What's the status with the base install step, do we still have show-stopper > bugs there? Yup, show-stopper is 123387 (busybox: wc reading from stdin is broken). Very annoying one is 123382 (kernel-image-2.2.20-idepci: need CONFIG_PACKET=y for boot-floppies dhcp-client, which I'll re-open and reassign to boot-floppies shortly). -David -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: [PATCH]? Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
I committed the newt hack, and filed a number of other bugs, the only complete show-stopper however is this one: http://bugs.debian.org/123387 busybox: wc reading from stdin is broken I believe this is what causes the Packages.gz corrupt problem that aph spoke of this weekend. -David -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: [PATCH]? Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
David Kimdon wrote: >Appears to work. The flashing screen went away, and I had a >choice of languages. I'll drop down and do a full install now and see >if anything else turns up. If all is cool and no one squeals I'll >commit the above patch. Okay, cool. We can always change it to use a more refined method later, like maybe actually making library reduction work for newt. But if your patch allows people to build working disk sets again, that's probably good enough for now. What's the status with the base install step, do we still have show-stopper bugs there? p. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
[PATCH]? Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
Looks like pb hit the nail on the head. Mon, Dec 10, 2001 at 01:53:21PM + wrote: > Oh, wait, this can't be good news: > > I: Retrieving >http://http.us.debian.org/debian/pool/main/n/newt/libnewt0_0.50.17-7.3_i386.deb > > I wonder if it's using that libnewt.so rather than the one from my host > system. yah, that looks like the problem. > Anyone want to try removing libnewt0 from EXTRACT_LIST_all and > seeing if that helps at all? I tried that but rootdisk.sh wasn't happy, it wants the libraries there. It sort of makes sense because archs that don't do library reduction won't have mklibs run, so they won't even get a copy of an un-reduced library, unless we want to add a dummy mklibs run that would do that. So . . . at the risk of proposing an inelegant hack: Index: rootdisk.sh === RCS file: /cvs/debian-boot/boot-floppies/rootdisk.sh,v retrieving revision 1.301 diff -u -r1.301 rootdisk.sh --- rootdisk.sh 2001/12/10 22:21:05 1.301 +++ rootdisk.sh 2001/12/11 08:21:06 @@ -218,8 +218,12 @@ info "downloading required packages from files $extract_list" debug `cat $extract_list` +if [ "$ROOT_IS_I18N" = 'true' ] ; then + PACKAGE_PATHS=$(grab_paths $(sed -e 's/ .*//; s/libnewt0/libnewt-utf8-0/' +$extract_list)) +else + PACKAGE_PATHS=$(grab_paths $(sed -e 's/ .*//' $extract_list)) +fi -PACKAGE_PATHS=$(grab_paths $(sed -e 's/ .*//' $extract_list)) (cat $extract_list | while read a b; do [ "$b" != "" ] || continue *shudder* Appears to work. The flashing screen went away, and I had a choice of languages. I'll drop down and do a full install now and see if anything else turns up. If all is cool and no one squeals I'll commit the above patch. -David -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
On Tue, Dec 11, 2001 at 12:23:56AM +0100, Claus Hindsgaul wrote: > I'm no programmer, but are you sure the file /release_notes.en exist?? When I > boot the b-f, the file /release_notes is only present *without* the ".en" > extension. this should be the case for non-i18n version only... -- Misha -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
I'm no programmer, but are you sure the file /release_notes.en exist?? When I boot the b-f, the file /release_notes is only present *without* the ".en" extension. Could this be the clue? Claus Adam De Carlo wrote: >causes the segmentation faults. I got a static strace on the >rootdisks and this is the strace ending: > >open("/release_notes.en", O_RDONLY) = 4 >read(4, "Software in "..., 1094) = 1094 >close(4)= 0 >select(4, [3], NULL, NULL, {0, 0}) = 0 (Timeout) >brk(0x8077000) = 0x8077000 >getpid()= 795 >ioctl(0, SNDCTL_TMR_TIMEBASE, {B38400 -opost isig -icanon -echo ...}) = 0 >--- SIGSEGV (Segmentation fault) --- > >If someone could track this down it would be very good. -- Claus Hindsgaul [EMAIL PROTECTED] (PGP-nøgle: http://www.image.dk/~claus_h/PGP.htm ) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
#include [EMAIL PROTECTED] wrote on Mon Dec 10, 2001 um 01:39:33PM: > Yes, they had different file sizes. But I can't figure out how that > happened and I'm beginning to suspect that maybe I got some files mixed > up along the way. I ran a new build with debugging turned on mklibs.py > and verified that it wasn't attempting to reduce libnewt, but I haven't > attempted to run the resulting binaries yet. I'm also completely baffled > by Eduard's mail where he says that installing libnewt-utf8-pic made the > difference. I wonder too. Looking what happens... ll /usr/lib/libnewt.so* lrwxrwxrwx1 root root 18 Nov 30 02:08 /usr/lib/libnewt.so -> libnewt.so.0.50.17 lrwxrwxrwx1 root root 18 Dec 9 02:14 /usr/lib/libnewt.so.0.50 -> libnewt.so.0.50.17 -rw-r--r--1 root root56392 Nov 28 18:02 /usr/lib/libnewt.so.0.50.17 removing libnewt-utf8-pic, building bfs, trying it (FLASHES!), mounting the root image: :/tmp/foo/lib> ll libnewt.so.0.50* -rw-r--r--2 root root56368 Dec 10 18:27 libnewt.so.0.50 -rw-r--r--2 root root56368 Dec 10 18:27 libnewt.so.0.50.17 - so the libs have beed damaged, IMHO. installing libnewt-utf8-pic, building, testing (OKAY!), mounting: :/tmp/bar/lib> ll libnewt.so.0.50* -rw-r--r--2 root root42948 Dec 10 18:39 libnewt.so.0.50 -rw-r--r--2 root root42948 Dec 10 18:39 libnewt.so.0.50.17 Gruss/Regards, Eduard. -- B.Gates: quality software :: R.McDonald: gourmet cuisine -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
Oh, wait, this can't be good news: I: Retrieving http://http.us.debian.org/debian/pool/main/n/newt/libnewt0_0.50.17-7.3_i386.deb I wonder if it's using that libnewt.so rather than the one from my host system. Anyone want to try removing libnewt0 from EXTRACT_LIST_all and seeing if that helps at all? p. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
In message <[EMAIL PROTECTED]>, Matt Zimmerman writes: >On Mon, Dec 10, 2001 at 11:07:01AM +, [EMAIL PROTECTED] wrote: > >> In message <[EMAIL PROTECTED]>, Adam Di Carlo writes: >> >This seems like a reasonable set of stuff for i18n builds, but it causes >> >the segmentation faults. I got a static strace on the rootdisks and this >> >is the strace ending: >> >> Yes, something is _still_ going wrong with libnewt. I don't understand >> what, yet, but I'm looking at it. >> >> I'd expect that libnewt.so on the root disk should be exactly the same as >> /usr/lib/libnewt.so on my build system (in absence of library reduction) >> yet this is not the case. And if I copy the one from the host into the >> root image, the segfault goes away. > >Did you check whether they are actually different? My system's libnewt.so >has exactly the same md5sum as the one on root.bin. Yes, they had different file sizes. But I can't figure out how that happened and I'm beginning to suspect that maybe I got some files mixed up along the way. I ran a new build with debugging turned on mklibs.py and verified that it wasn't attempting to reduce libnewt, but I haven't attempted to run the resulting binaries yet. I'm also completely baffled by Eduard's mail where he says that installing libnewt-utf8-pic made the difference. p. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
On Mon, Dec 10, 2001 at 11:07:01AM +, [EMAIL PROTECTED] wrote: > In message <[EMAIL PROTECTED]>, Adam Di Carlo writes: > >This seems like a reasonable set of stuff for i18n builds, but it causes > >the segmentation faults. I got a static strace on the rootdisks and this > >is the strace ending: > > Yes, something is _still_ going wrong with libnewt. I don't understand > what, yet, but I'm looking at it. > > I'd expect that libnewt.so on the root disk should be exactly the same as > /usr/lib/libnewt.so on my build system (in absence of library reduction) > yet this is not the case. And if I copy the one from the host into the > root image, the segfault goes away. Did you check whether they are actually different? My system's libnewt.so has exactly the same md5sum as the one on root.bin. -- - mdz -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
In message <[EMAIL PROTECTED]>, Adam Di Carlo writes: >This seems like a reasonable set of stuff for i18n builds, but it >causes the segmentation faults. I got a static strace on the >rootdisks and this is the strace ending: Yes, something is _still_ going wrong with libnewt. I don't understand what, yet, but I'm looking at it. I'd expect that libnewt.so on the root disk should be exactly the same as /usr/lib/libnewt.so on my build system (in absence of library reduction) yet this is not the case. And if I copy the one from the host into the root image, the segfault goes away. p. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
#include Adam Di Carlo wrote on Sun Dec 09, 2001 um 10:23:43PM: > > > I'm annoyed that we cannot figure out a set of newt and slang libs > (including the -pic stuff) which can be used to reliably build this > stuff. As such I'm turning off i18n for now. > > Here's what I had installed on the breaking builds: > > libnewt-dev 0.50.17-7.3 > libnewt-utf8-0 0.50.17-7.3 > slang1 1.4.4-6 > slang1-utf8 1.4.4-6 > slang1-utf8-dev 1.4.4-6 > slang1-utf8-pic 1.4.4-6 Sure? make check complains if libnewt-utf8-pic is installed. If I remove it, I get the flashing-screen-syndrome. If it is installed and the build done properly, lang-chooser DOES WORK. Weird. pn libnews-newsrc (no description available) pn libnews-nntpcl (no description available) pn libnews-scan-p (no description available) ii libnewt-dev0.50.17-7.3Developer's toolkit for newt windowing libra pn libnewt-perl(no description available) pn libnewt-pic (no description available) pn libnewt-utf8-0 (no description available) ii libnewt-utf8-p 0.50.17-7.3newt shared library subset kit, with utf8 su ii libnewt0 0.50.17-7.3Not Erik's Windowing Toolkit - text mode win ii slang1 1.4.4-6The S-Lang programming library - runtime ver pn slang1-dev (no description available) pn slang1-ja (no description available) pn slang1-ja-dev (no description available) pn slang1-pic (no description available) rc slang1-utf81.4.4-6The S-Lang programming library with utf8 sup ii slang1-utf8-de 1.4.4-6The S-Lang programming library, development ii slang1-utf8-pi 1.4.4-6The S-Lang programming library, shared libra Gruss/Regards, Eduard. -- 16:46 sorry fuer die andauernden rejoins 16:46 -!- Molle [~[EMAIL PROTECTED]] has quit [leaving] -- #debian.de -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
Mon, Dec 10, 2001 at 01:39:04AM -0500 wrote: > On Sun, Dec 09, 2001 at 09:54:14PM -0800, David Kimdon wrote: > > > ifeq ($(LC),true) > > -CFLAGS += -DUSE_LANGUAGE_CHOOSER > > +CFLAGS += -DUSE_LANGUAGE_CHOOSE -DUTF8 > ^ Is this intentional? yah, my mistake, all this patch does is turn off langauage choser, funny, it makes the disk behave better, who would have thought? Back to the drawing board. Thanks for pointing that out. -David -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
On Sun, Dec 09, 2001 at 09:54:14PM -0800, David Kimdon wrote: > ifeq ($(LC),true) > -CFLAGS += -DUSE_LANGUAGE_CHOOSER > +CFLAGS += -DUSE_LANGUAGE_CHOOSE -DUTF8 ^ Is this intentional? -- - mdz -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
I'm getting better behavior with this patch. It is still rebooting, but at least it is telling me that there was an error loading some messages, and I can read it, rather than just a flashing screen. Index: utilities/dbootstrap/Makefile === RCS file: /cvs/debian-boot/boot-floppies/utilities/dbootstrap/Makefile,v retrieving revision 1.102 diff -u -r1.102 Makefile --- utilities/dbootstrap/Makefile 2001/12/03 02:40:45 1.102 +++ utilities/dbootstrap/Makefile 2001/12/10 05:22:07 @@ -65,7 +65,7 @@ endif ifeq ($(LC),true) -CFLAGS += -DUSE_LANGUAGE_CHOOSER +CFLAGS += -DUSE_LANGUAGE_CHOOSE -DUTF8 endif ifdef KBD I think the culprit may be somewhere in slang.h. This is only a half-baked idea, but if anyone else is working on it this evening, perhaps they'd like to try the above patch as well. -David -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: forget LANG_CHOOSER for 3.0.18 (was Re: 3.0.18 testing)
On Sun, Dec 09, 2001 at 10:23:43PM -0500, Adam Di Carlo wrote: > This seems like a reasonable set of stuff for i18n builds, but it > causes the segmentation faults. I got a static strace on the > rootdisks and this is the strace ending: > > open("/release_notes.en", O_RDONLY) = 4 > read(4, "Software in "..., 1094) = 1094 > close(4)= 0 > select(4, [3], NULL, NULL, {0, 0}) = 0 (Timeout) > brk(0x8077000) = 0x8077000 > getpid()= 795 > ioctl(0, SNDCTL_TMR_TIMEBASE, {B38400 -opost isig -icanon -echo ...}) = 0 > --- SIGSEGV (Segmentation fault) --- > > > If someone could track this down it would be very good. The segfault occurs here: (gdb) bt #0 0x4007bb2f in malloc () from /lib/libc.so.6 #1 0x4007b1b4 in malloc () from /lib/libc.so.6 #2 0x4009ed97 in ttyname () from /lib/libc.so.6 #3 0x4001c4e9 in newtButton () from /lib/libnewt.so.0.50 #4 0x4001d76b in newtFormRun () from /lib/libnewt.so.0.50 #5 0x4001d521 in newtRunForm () from /lib/libnewt.so.0.50 #6 0x0804d502 in strcpy () Cannot access memory at address 0x16 However, the libnewt.so.0.50 in root.bin is exactly the same as the one on my system, so I don't think that's the problem. If I use the same dbootstrap binary, but link it an runtime with the libslang.so.1 and libm.so.6 from my working system, it works. So it looks like probably the slang library is to blame. -- - mdz -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]