Re: pbuilder help (bug 334877)
Hi, Well, this build takes a long time (at least 45 minutes from start to failure). I'm getting tired of trying over and over again; it took me about four hours of compiling to prove certainly that Blars suggestion wouldn't work. Can you give me a hook script that will give an interactive shell? I tried the obvious: #!/bin/sh /bin/bash Heh, I was very slow to read this mail; but it's available in /usr/share/doc/pbuilder/examples/C10shell which looks like: #!/bin/bash # invoke shell if build fails. /bin/bash /dev/tty /dev/tty regards, junichi -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Blars Blarson [EMAIL PROTECTED] writes: In article [EMAIL PROTECTED] [EMAIL PROTECTED] writes: I'm trying to solve bug 304932/334877. I can reproduce the build failure using pbuilder, but not when I build on my own system directly. I would like to do the pbuilder build and then examine the failing filesystem, but pbuilder always deletes the build directory, and the manual gives no clear indication of how to prevent this. --debug says that it only avoids cleanup in update and create. Use: pbuilder login sed -i~ -e 's/#//g' /etc/apt/sources.list cd /tmp/build apt-get build-dep $package apt-get source $package cd $package-ver dpkg-buildpackage -us -uc Unfortunately, when I build that way, the build succeeds. Hence my question: I want to build with pbuilder build, and be able to examine the filesystem at the point it crashes. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
On Friday, 21 October 2005 07:26, Thomas Bushnell BSG wrote: I'm trying to solve bug 304932/334877. I can reproduce the build failure using pbuilder, but not when I build on my own system directly. I would like to do the pbuilder build and then examine the failing filesystem, but pbuilder always deletes the build directory, and the manual gives no clear indication of how to prevent this. --debug says that it only avoids cleanup in update and create. Use pbuilder hooks for that, they allow great flexibility, for example, examining the filesystem after failing or using ccache to speed up build time. Best regards -- Isaac Clerencia at Warp Networks, http://www.warp.es Work: [EMAIL PROTECTED] | Debian: [EMAIL PROTECTED] pgpISLWXgsW77.pgp Description: PGP signature
Re: pbuilder help (bug 334877)
Isaac Clerencia [EMAIL PROTECTED] writes: On Friday, 21 October 2005 07:26, Thomas Bushnell BSG wrote: I'm trying to solve bug 304932/334877. I can reproduce the build failure using pbuilder, but not when I build on my own system directly. I would like to do the pbuilder build and then examine the failing filesystem, but pbuilder always deletes the build directory, and the manual gives no clear indication of how to prevent this. --debug says that it only avoids cleanup in update and create. Use pbuilder hooks for that, they allow great flexibility, for example, examining the filesystem after failing or using ccache to speed up build time. Well, this build takes a long time (at least 45 minutes from start to failure). I'm getting tired of trying over and over again; it took me about four hours of compiling to prove certainly that Blars suggestion wouldn't work. Can you give me a hook script that will give an interactive shell? I tried the obvious: #!/bin/sh /bin/bash which did not succeed; it ran the script, but apparently pbuilder doesn't pass stdin/stdout down to the scripts. Surely it's not too hard to have a don't clean option; after all, it exists for the login and exec modes. Regardless, I'm back to there is no way to debug this problem. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
On Saturday, 22 October 2005 00:36, Thomas Bushnell BSG wrote: Isaac Clerencia [EMAIL PROTECTED] writes: On Friday, 21 October 2005 07:26, Thomas Bushnell BSG wrote: I'm trying to solve bug 304932/334877. I can reproduce the build failure using pbuilder, but not when I build on my own system directly. I would like to do the pbuilder build and then examine the failing filesystem, but pbuilder always deletes the build directory, and the manual gives no clear indication of how to prevent this. --debug says that it only avoids cleanup in update and create. Use pbuilder hooks for that, they allow great flexibility, for example, examining the filesystem after failing or using ccache to speed up build time. Well, this build takes a long time (at least 45 minutes from start to failure). I'm getting tired of trying over and over again; it took me about four hours of compiling to prove certainly that Blars suggestion wouldn't work. I've an ugly hack to get ccache working inside the pbuilder, that saves lots of build time: [EMAIL PROTECTED]:~/debian/pbuilder-hooks$ cat A01sh #!/bin/sh echo Setting up ccache apt-get install ccache cd /usr/lib/ccache for c in * do ln -s /usr/bin/ccache /usr/sbin/$c done ln -s /var/cache/pbuilder/ccache /root/.ccache and I run pbuilder as: pdebuild -- --bindmounts /var/cache/pbuilder/ccache --hookdir ~/debian/pbuilder-hooks/ which did not succeed; it ran the script, but apparently pbuilder doesn't pass stdin/stdout down to the scripts. Well, I tried that and didn't managed to get it working so I just put there a sleep 999 so I can chroot into the pbuilder build dir and debug the problem :) -- Isaac Clerencia at Warp Networks, http://www.warp.es Work: [EMAIL PROTECTED] | Debian: [EMAIL PROTECTED] pgpYceYncWvMD.pgp Description: PGP signature
Re: pbuilder help (bug 334877)
Isaac Clerencia [EMAIL PROTECTED] writes: I've an ugly hack to get ccache working inside the pbuilder, that saves lots of build time. Thanks, but the big build time for lilypond is mostly consumed with tracing fonts, not compiling C code. :( Well, I tried that and didn't managed to get it working so I just put there a sleep 999 so I can chroot into the pbuilder build dir and debug the problem :) Blech. So pbuilder, for all its charms, is missing one of the single most useful features it could have. I'll do your hack though; thanks for the suggestion. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
I demand that Thomas Bushnell BSG may or may not have written... Isaac Clerencia [EMAIL PROTECTED] writes: I've an ugly hack to get ccache working inside the pbuilder, that saves lots of build time. Thanks, but the big build time for lilypond is mostly consumed with tracing fonts, not compiling C code. :( A uuencoded tarball of the generated files would appear to be useful here. (You'll probably want tar's -m option when unpacking.) Or have you already tried this? [snip] -- | Darren Salt | linux (or ds) at | nr. Ashington, | sarge,| youmustbejoking | Northumberland | RISC OS | demon co uk | Toon Army | URL:http://www.youmustbejoking.demon.co.uk/ (PGP 2.6, GPG keys) But I don't like Spam!!! -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Darren Salt [EMAIL PROTECTED] writes: A uuencoded tarball of the generated files would appear to be useful here. (You'll probably want tar's -m option when unpacking.) Well, I'm now closer in. The first invocation of lilypond from within the build fails. But it fails silently. The failing build prints (at the key place): cd ./out /tmp/buildd/lilypond-2.6.3/lily/out/lilypond --verbose /tmp/buildd/lilypond-2.6.3/ly/generate-documentation LILYPOND_DATADIR=/usr/share/lilypond/2.6.3 LILYPONDPREFIX=/tmp/buildd/lilypond-2.6.3/share/lilypond/2.6.3 LOCALEDIR=/usr/share/locale Effective prefix: /tmp/buildd/lilypond-2.6.3/share/lilypond/2.6.3 Initializing FontConfig... adding font directory: /tmp/buildd/lilypond-2.6.3/share/lilypond/2.6.3/fonts/otf/ adding font directory: /tmp/buildd/lilypond-2.6.3/share/lilypond/2.6.3/fonts/type1/ adding font directory: /tmp/buildd/lilypond-2.6.3/share/lilypond/2.6.3/fonts/svg/rm -f ./out/lilypond-internals.nexi ln ./out/lilypond-internals.texi ./out/lilypond-internals.nexi ln: accessing `./out/lilypond-internals.texi': No such file or directory make[3]: [out/lilypond-internals.texi] Error 1 (ignored) Now the lines from rm -f ./out/lilypond-internals-nexi are from the invoking makefile, as is the first line (cd ./out...) The LILYPOND_... output, the Effective prefix schtick, and the FontConfig output are all from lilypond (which was invoked with --verbose to show this). Notice that there is no newline between lilypond's last output (the third adding font directory bit) and the next output from the invoking makefile. In a correct invocation, there would be threelines to follow: a line Processing `/tmp/buildd/lilypond-2.6.3/ly/generate-documentation.ly' and a line starting Parsing... which is the output of the Scheme program that needs to run here, and then: Writing lilypond-internals.texi...]] Ok, so lilypond is failing. But dammitall, I can't get it to fail ever else. If I run this command myself after the failure, it works fine. Likewise if I invoke make, or if I clean the directory and go up and do debian/rules build. Clearly something about the buildd dynamic environment is *different* from what I get if I just enter and do it myself, and that difference causes the generated lilypond to fail. Any suggestions for how I can poke at this further? Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Thomas Bushnell BSG wrote: Ok, so lilypond is failing. But dammitall, I can't get it to fail ever else. If I run this command myself after the failure, it works fine. Likewise if I invoke make, or if I clean the directory and go up and do debian/rules build. Clearly something about the buildd dynamic environment is *different* from what I get if I just enter and do it myself, and that difference causes the generated lilypond to fail. Quick random guess: Could pbuilder be providing either no stdin, or something silly for stdin? Other than that, check the environment... should be fairly easy to throw an 'env' call in the makefile. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Anthony DeRobertis [EMAIL PROTECTED] writes: Clearly something about the buildd dynamic environment is *different* from what I get if I just enter and do it myself, and that difference causes the generated lilypond to fail. Quick random guess: Could pbuilder be providing either no stdin, or something silly for stdin? It's possible, but the command invoked doesn't (shouldn't?!) read from stdin. I'll check this out... Other than that, check the environment... should be fairly easy to throw an 'env' call in the makefile. Yep, I have a build proceeding now with the equivalent in a pbuilder hook. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Thomas Bushnell BSG [EMAIL PROTECTED] writes: Quick random guess: Could pbuilder be providing either no stdin, or something silly for stdin? It's possible, but the command invoked doesn't (shouldn't?!) read from stdin. I'll check this out... Good golly, Miss Molly, that's it. It does indeed blow chunks if the input is /dev/null (whether within a chroot or just a normal native build). Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Thomas Bushnell BSG wrote: Good golly, Miss Molly, that's it. It does indeed blow chunks if the input is /dev/null (whether within a chroot or just a normal native build). Heh. Glad that helped. Took a wild guess from your previous message about hooks not getting stdin. Now, you just get the fun of figuring out why -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Anthony DeRobertis [EMAIL PROTECTED] writes: Thomas Bushnell BSG wrote: Good golly, Miss Molly, that's it. It does indeed blow chunks if the input is /dev/null (whether within a chroot or just a normal native build). Heh. Glad that helped. Took a wild guess from your previous message about hooks not getting stdin. Now, you just get the fun of figuring out why Well, I've reported it upstream. They're pretty good about helping out, and understand the internal details far better than I. :) Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
Thomas Bushnell BSG [EMAIL PROTECTED] writes: Anthony DeRobertis [EMAIL PROTECTED] writes: Thomas Bushnell BSG wrote: Good golly, Miss Molly, that's it. It does indeed blow chunks if the input is /dev/null (whether within a chroot or just a normal native build). Heh. Glad that helped. Took a wild guess from your previous message about hooks not getting stdin. Now, you just get the fun of figuring out why Well, I've reported it upstream. They're pretty good about helping out, and understand the internal details far better than I. :) Lilypond thinks that if input is not a terminal, we must be running under a GUI. At that point, its behavior starts changing in all sorts of subtle ways (part of the output now goes to a log file instead of stdout, random chdir happens partway through, etc.). Sigh. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: pbuilder help (bug 334877)
In article [EMAIL PROTECTED] [EMAIL PROTECTED] writes: I'm trying to solve bug 304932/334877. I can reproduce the build failure using pbuilder, but not when I build on my own system directly. I would like to do the pbuilder build and then examine the failing filesystem, but pbuilder always deletes the build directory, and the manual gives no clear indication of how to prevent this. --debug says that it only avoids cleanup in update and create. Use: pbuilder login sed -i~ -e 's/#//g' /etc/apt/sources.list cd /tmp/build apt-get build-dep $package apt-get source $package cd $package-ver dpkg-buildpackage -us -uc -- Blars Blarson [EMAIL PROTECTED] http://www.blars.org/blars.html With Microsoft, failure is not an option. It is a standard feature. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]