Source: dcmtk
Version: 3.6.7-9
Severity: serious
Tags: ftbfs
Justification: fails to build from source (but built successfully in the past) X-Debbugs-Cc: sramacher@debian.org

https://buildd.debian.org/status/fetch.php?pkg=dcmtk&arch=armel&ver=3.6.7-9%2Bb1&stamp=1704494616&raw=0

[ 90%] Building CXX object dcmpstat/libsrc/CMakeFiles/dcmpstat.dir/dcmpstat.cc.o
cd /<<PKGBUILDDIR>>/obj-arm-linux-gnueabi/dcmpstat/libsrc && /usr/bin/c++ -DDCMTK_BUILD_IN_PROGRESS -DDCMTK_HAVE_POLL=1 -DUSE_NULL_SAFE_OFSTRING -D_REENTRANT -Ddcmpstat_EXPORTS -I/<<PKGBUILDDIR>>/obj-arm-linux-gnueabi/config/include -I/<<PKGBUILDDIR>>/
ofstd/include -I/<<PKGBUILDDIR>>/oflog/include -I/<<PKGBUILDDIR>>/dcmdata/include -I/<<PKGBUILDDIR>>/dcmimgle/include -I/<<PKGBUILDDIR>>/dcmimage/include -I/<<PKGBUILDDIR>>/dcmjpeg/include -I/<<PKGBUILDDIR>>/dcmjpls/include -I/<<PKGBUILDDIR>>/dcmtls/
include -I/<<PKGBUILDDIR>>/dcmnet/include -I/<<PKGBUILDDIR>>/dcmsr/include -I/<<PKGBUILDDIR>>/dcmsign/include -I/<<PKGBUILDDIR>>/dcmwlm/include -I/<<PKGBUILDDIR>>/dcmqrdb/include -I/<<PKGBUILDDIR>>/dcmpstat/include -I/<<PKGBUILDDIR>>/dcmrt/include -I/<<
PKGBUILDDIR>>/dcmiod/include -I/<<PKGBUILDDIR>>/dcmfg/include -I/<<PKGBUILDDIR>>/dcmseg/include -I/<<PKGBUILDDIR>>/dcmtract/include -I/<<PKGBUILDDIR>>/dcmpmap/include -I/<<PKGBUILDDIR>>/dcmect/include -I/usr/include/libxml2 -g -O2 -ffile-prefix-map=/<<
PKGBUILDDIR>>=. -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -DENABLE_DCMJPLS_INTERLEAVE_NONE -Wdate-time -D_FORTIFY_SOURCE=2 -fvisibility=hidden -D_XOPEN_SOURCE_EXTENDED -D_XOPEN_SOURCE=500 -D_BSD_SOURCE -D_DEFAULT_
SOURCE -D_BSD_COMPAT -D_OSF_SOURCE -D_POSIX_C_SOURCE=199506L -std=gnu++17 -fPIC -Wall -MD -MT dcmpstat/libsrc/CMakeFiles/dcmpstat.dir/dcmpstat.cc.o -MF CMakeFiles/dcmpstat.dir/dcmpstat.cc.o.d -o CMakeFiles/dcmpstat.dir/dcmpstat.cc.o -c /<<PKGBUILDDIR>>/
dcmpstat/libsrc/dcmpstat.cc
/tmp/ccm0eYhx.s: Assembler messages:
/tmp/ccm0eYhx.s:537: Error: bad immediate value for offset (4100)
make[4]: *** [dcmrt/tests/CMakeFiles/drttest.dir/build.make:79: dcmrt/tests/CMakeFiles/drttest.dir/drttest.cc.o] Error 1
make[4]: Leaving directory '/<<PKGBUILDDIR>>/obj-arm-linux-gnueabi'
make[3]: *** [CMakeFiles/Makefile2:5064: dcmrt/tests/CMakeFiles/drttest.dir/all] Error 2
make[3]: *** Waiting for unfinished jobs....
[ 90%] Building CXX object dcmpstat/libsrc/CMakeFiles/dcmpstat.dir/dviface.cc.o

Cheers
--
Sebastian Ramacher

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Control: fixed -1 3.6.8-3

On Sat, Jan 13, 2024 at 9:42 PM Emanuele Rocca <ema@debian.org> wrote:

Control: user -1 debian-arm@lists.debian.org
Control: usertag -1 + 32bit-stackclash

Hi,

On Fri, Jan 05, 2024 at 11:45:28PM +0100, Sebastian Ramacher wrote:

/tmp/ccm0eYhx.s: Assembler messages:
/tmp/ccm0eYhx.s:537: Error: bad immediate value for offset (4100)

This is caused by stack-clash-protection on armel and a workaround is in version 3.6.8-3 currently in experimental, see: https://tracker.debian.org/news/1494233/accepted-dcmtk-368-3-source-into-experimental/

We should downgrade the severity to minor once the fix enters unstable, but keep the bug open as this seems to be an interesting case of stack-clash-protection malfunctioning on 32bit arm to further look into.

Bit lost here... I do not see the bug reported against GCC-13 package.

In the end do you want me to upload a patched 3.6.7 or is it ok to
wait for transition ?

Thanks!

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Processing control commands:

fixed -1 3.6.8-3

Bug #1060104 [src:dcmtk] dcmtk: FTBFS on armel: Error: bad immediate value for offset (4100)
Marked as fixed in versions dcmtk/3.6.8-3.

--
1060104: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1060104
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Mathieu,

On 2024-01-14 11:24, Mathieu Malaterre wrote:

On Sat, Jan 13, 2024 at 9:42 PM Emanuele Rocca <ema@debian.org> wrote:

We should downgrade the severity to minor once the fix enters unstable, but keep the bug open as this seems to be an interesting case of stack-clash-protection malfunctioning on 32bit arm to further look into.

Bit lost here... I do not see the bug reported against GCC-13 package.

Sorry for the lack of context. Out of all packages in the archive, a few
(less than 10 it seems, rebuilds still ongoing) fail to build on armhf
or armel with the stack-clash-protection flag on, but build successfully without it. It's not entirely clear why this is the case yet, but we
know that disabling the flag fixes the issue. For this reason I would
suggest to disable stackclash on the armel build of dcmtk (just like you
did in experimental) to make sure the package builds properly again, but
keep #1060104 open at a lower severity so that we don't lose track of
this.

In the end do you want me to upload a patched 3.6.7 or is it ok to
wait for transition ?

Up to you really, no rush on my side.

Thanks,
Emanuele

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Processing control commands:

severity -1 important

Bug #1060104 [src:dcmtk] dcmtk: FTBFS on armel: Error: bad immediate value for offset (4100)
Severity set to 'important' from 'serious'

--
1060104: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1060104
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Control: severity -1 important

On Mon, Jan 15, 2024 at 1:49 PM Emanuele Rocca <ema@debian.org> wrote:
[...]

For this reason I would
suggest to disable stackclash on the armel build of dcmtk (just like you
did in experimental) to make sure the package builds properly again, but
keep #1060104 open at a lower severity so that we don't lose track of
this.

Done ! Thanks

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Processing control commands:

severity -1 serious

Bug #1060104 [src:dcmtk] dcmtk: FTBFS on armel: Error: bad immediate value for offset (4100)
Severity set to 'serious' from 'important'

--
1060104: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1060104
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Dear Emmanuele and Mathieu,

On 3/19/24 08:51, Mathieu Malaterre wrote:

On Tue, Mar 19, 2024 at 8:44 AM Emanuele Rocca <ema@debian.org> wrote:

Hi,

On 2024-03-19 06:24, Sébastien Jodogne wrote:

Because of bug #1060104, a large majority of the packages related to
medical imaging have just disappeared from Debian Unstable.

But, if I correctly understand #1060104, it is specific to one single
platform (armel).

Indeed, and there is a simple fix too, which has been uploaded to
experimental only so far:
https://salsa.debian.org/med-team/dcmtk/-/commit/42583dfe9fd344a63cdbc278268d4176d4a22ec4

Mathieu (or someone else from debian-med), could you please apply that
to unstable as well? It seems that with the current state of unstable
the transition will take a while anyways.

I will be away from any Debian tasks for at least another month unfortunately. The patch was suggested by an armel porter so I believe
this is the right thing to do.

Worth pointing out that right now dcmtk cannot be built in sid/armel due
to a missing build depend, namely graphviz. It seems worth applying the
fix to unstable anyways so that it does not fall through the cracks, and
we can schedule a binNMU later when graphviz is available again.

I could try to upload this patch by myself to unstable. Unfortunately,
I'm not an uploader of the dcmtk package, so an intervention from a
Debian Developer is required to give me the proper access rights: https://tracker.debian.org/pkg/dcmtk

Kind Regards,
Sébastien-

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Dear all,

Indeed, and there is a simple fix too, which has been uploaded to experimental only so far:

yes, there is a simple fix (that disables the error), but there is also a more fundamental fix that (hopefully) solves the issue at its core:

https://git.dcmtk.org/?p=dcmtk.git;a=commit;h=bdc18c85b7fca130464dd745ea3efef3c9b8b94e

Furthermore, it feels kind of strange that the DCMTK package was about to be removed because a rather insignificant test tool causes problems on a single platform, and this test tool is even not part of any Debian package (see "debian/rules" file):

override_dh_install-arch:
[...]
#remove test binaries
[...]
rm ./debian/dcmtk/usr/bin/drttest
[...]

Regards,
Jörg
--
Dr. Joerg Riesmeier, Etzhorner Weg 19, 26125 Oldenburg, Germany
E-Mail: dicom@jriesmeier.com, Homepage: www.jriesmeier.com

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Emanuele.

Am Tue, Mar 19, 2024 at 12:16:16PM +0100 schrieb Emanuele Rocca:

I've uploaded a NMU to DELAYED/2: https://bugs.debian.org/1067147

Thanks a lot for your attempt to help. In principle we have a team wide
low threshold NMU - so undelayed NMUs are fine. The kind of race
condition in uploads might imply that some ping in advance might be
helpful to avoid duplicated work.

With those changes dcmtk builds fine on both armel and armhf. I've
dropped the graphviz build-depend on those arches too, it can of course
be reintroduced once graphviz become installable again.

Meanwhile your patch is imported and was uploaded by Michael - thanks to
all who helped solving this.

Kind regards
Andreas.

--
http://fam-tille.de

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)