4:42 a.m.
I maintain a package written in Pascal, which uses fpc for compiling. I noticed that
recently, koji builds started failing on i686 and armv7hl due to the find-debuginfo script
failing to, well, extract the debuginfo.
Here's a link to a failed koji build (mass-rebuild by releng):
https://koji.fedoraproject.org/koji/taskinfo?taskID=20981009
On x86_64 and ppc64, the package builds fine. On i686 and armv7hl, this happens:
extracting debug info from
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
Stabs debuginfo not supported:
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
gdb-add-index: No index was created for
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
gdb-add-index: [Was there no debuginfo? Was there already an index?]
[ snip ]
RPM build errors:
error: Empty %files file
/builddir/build/BUILD/LD25-e5ecbe39b719f12a1268bcc641eae9ba364221c9/debugsourcefiles.list
I thought this may be a bug with the package, but then I saw that koji builds for a new
fpc-compiled package that I've been working on lately (not yet submitted for review)
suffer the same problem. This one is worse, even, since find-debuginfo fails with a
coredump. https://koji.fedoraproject.org/koji/taskinfo?taskID=21087076
extracting debug info from
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-qt
extracting debug info from
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-gtk
Stabs debuginfo not supported:
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-qt
Stabs debuginfo not supported:
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-gtk
dwz: dwz.c:1984: checksum_die: Assertion `((!op_multifile && !rd_multifile
&& !fi_multifile) || cu != die_cu (ref)) && (!op_multifile ||
cu->cu_chunk == die_cu (ref)->cu_chunk)' failed.
/usr/lib/rpm/find-debuginfo.sh: line 518: 8822 Aborted (core dumped) dwz
$dwz_opts ${dwz_files[@]}
I checked the build & updates status for fpc and there haven't been any updates to
the package lately, so this definitely isn't a regression in the compiler.
Have there been any recent changes to find-debuginfo that may be causing this?
5:58 a.m.
New subject: find-debuginfo fails for fpc-compiled software on i686 and
armv7hl
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
On Mon, 2017-08-07 at 09:42 +0000, Artur Iwicki wrote:
I maintain a package written in Pascal, which uses fpc for
compiling.
I noticed that recently, koji builds started failing on i686 and
armv7hl due to the find-debuginfo script failing to, well, extract
the debuginfo.
It is problem of FPC, see
https://bugzilla.redhat.com/show_bug.cgi?id=1475223.
Here's a link to a failed koji build (mass-rebuild by releng):
https://koji.fedoraproject.org/koji/taskinfo?taskID=20981009
On x86_64 and ppc64, the package builds fine. On i686 and armv7hl,
this happens:
> extracting debug info from /builddir/build/BUILDROOT/colorful-1.3-
> 3.fc27.arm/usr/bin/colorful
> Stabs debuginfo not supported: /builddir/build/BUILDROOT/colorful-
> 1.3-3.fc27.arm/usr/bin/colorful
> gdb-add-index: No index was created for
> /builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
> gdb-add-index: [Was there no debuginfo? Was there already an
> index?]
[ snip ]
> RPM build errors:
> error: Empty %files file /builddir/build/BUILD/LD25-
> e5ecbe39b719f12a1268bcc641eae9ba364221c9/debugsourcefiles.list
I thought this may be a bug with the package, but then I saw that
koji builds for a new fpc-compiled package that I've been working on
lately (not yet submitted for review) suffer the same problem. This
one is worse, even, since find-debuginfo fails with a coredump. https
://koji.fedoraproject.org/koji/taskinfo?taskID=21087076
> extracting debug info from /builddir/build/BUILDROOT/peazip-6.4.1-
> 3.fc27.i386/usr/bin/peazip-qt
> extracting debug info from /builddir/build/BUILDROOT/peazip-6.4.1-
> 3.fc27.i386/usr/bin/peazip-gtk
> Stabs debuginfo not supported: /builddir/build/BUILDROOT/peazip-
> 6.4.1-3.fc27.i386/usr/bin/peazip-qt
> Stabs debuginfo not supported: /builddir/build/BUILDROOT/peazip-
> 6.4.1-3.fc27.i386/usr/bin/peazip-gtk
> dwz: dwz.c:1984: checksum_die: Assertion `((!op_multifile &&
> !rd_multifile && !fi_multifile) || cu != die_cu (ref)) &&
> (!op_multifile || cu->cu_chunk == die_cu (ref)->cu_chunk)' failed.
> /usr/lib/rpm/find-debuginfo.sh: line 518: 8822
> Aborted (core dumped) dwz $dwz_opts ${dwz_files[@]}
I checked the build & updates status for fpc and there haven't been
any updates to the package lately, so this definitely isn't a
regression in the compiler.
Have there been any recent changes to find-debuginfo that may be
causing this?
_______________________________________________
devel mailing list -- devel(a)lists.fedoraproject.org
To unsubscribe send an email to devel-leave(a)lists.fedoraproject.org
- --
- -Igor Gnatenko
-----BEGIN PGP SIGNATURE-----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=sSy9
-----END PGP SIGNATURE-----
6:42 a.m.
It is problem of FPC, see
https://bugzilla.redhat.com/show_bug.cgi?id=1475223.
Thank you. Very insightful.
6:16 a.m.
New subject: find-debuginfo fails for fpc-compiled software on i686 and
armv7hl
On Mon, 2017-08-07 at 09:42 +0000, Artur Iwicki wrote:
I maintain a package written in Pascal, which uses fpc for compiling.
I noticed that recently, koji builds started failing on i686 and armv7hl due to the
find-debuginfo script failing to, well, extract the debuginfo.
Here's a link to a failed koji build (mass-rebuild by releng):
https://koji.fedoraproject.org/koji/taskinfo?taskID=20981009
On x86_64 and ppc64, the package builds fine. On i686 and armv7hl, this happens:
> extracting debug info from
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
> Stabs debuginfo not supported:
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
> gdb-add-index: No index was created for
/builddir/build/BUILDROOT/colorful-1.3-3.fc27.arm/usr/bin/colorful
> gdb-add-index: [Was there no debuginfo? Was there already an index?]
[ snip ]
>RPM build errors:
>error: Empty %files file
/builddir/build/BUILD/LD25-e5ecbe39b719f12a1268bcc641eae9ba364221c9/debugsourcefiles.list
This is still being discussed upstream:
http://lists.rpm.org/pipermail/rpm-maint/2017-August/006245.html
But I think the consensus is that it is a packaging bug.
It means that although you package contains native binaries there is no
correct DWARF produced so no corresponding source files have been found.
This is usually caused by a missing -g in the build flags or the native
compiler (fpc in your case I assume) not producing correct DWARF
debugging information.
I thought this may be a bug with the package, but then I saw that
koji builds for a new fpc-compiled package that I've been working on lately (not yet
submitted for review) suffer the same problem. This one is worse, even, since
find-debuginfo fails with a coredump.
https://koji.fedoraproject.org/koji/taskinfo?taskID=21087076
>extracting debug info from
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-qt
>extracting debug info from
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-gtk
>Stabs debuginfo not supported:
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-qt
>Stabs debuginfo not supported:
/builddir/build/BUILDROOT/peazip-6.4.1-3.fc27.i386/usr/bin/peazip-gtk
That is definitely a bug in your package or the fpc compiler.
Stabs is a ancient debuginfo format that isn't supported by any/most
modern tools. You should definitely see if you can figure out how that
happened and how you can produce normal DWARF debuginfo.
>dwz: dwz.c:1984: checksum_die: Assertion `((!op_multifile
&& !rd_multifile && !fi_multifile) || cu != die_cu (ref)) &&
(!op_multifile || cu->cu_chunk == die_cu (ref)->cu_chunk)' failed.
>/usr/lib/rpm/find-debuginfo.sh: line 518: 8822 Aborted (core dumped)
dwz $dwz_opts ${dwz_files[@]}
I am not sure exactly how this happened. It happens when handling a
Dwarf DW_FORM_ref_addr construct. You might have to report a bug report
against dwz with the files so we can track this down.
But note that this doesn't abort the build (maybe it should, but it
currently doesn't - it just means your DWARF debuginfo won't be
compressed).
The reason your build failed is the first issue of missing/incorrect
DWARF generation so that no source files could be found.
Cheers,
Mark
6:41 a.m.
Stabs is a ancient debuginfo format that isn't supported by
any/most modern tools.
Hm. But it seemed to worked fine as recently as July 25.
You should definitely see if [...] you can produce normal DWARF
debuginfo.
I edited the makefile to use a different compiler switch, and with DWARF
debuginfo, the koji build completes just fine:
https://koji.fedoraproject.org/koji/taskinfo?taskID=21089412
Either way, I've read the Bugzilla ticket linked to by Igor and it seems that on i686,
fpc defaults to stabs debuginfo, so using the "generate DWARF debuginfo"
compiler switch is (currently) the preferred solution to the problem.
2461
days inactive
2461
days old
4 comments
3 participants
participants (3)
-
Artur Iwicki -
Igor Gnatenko -
Mark Wielaard