just to note that the libffi issues are back.
I believe Adrian put in some patched version with a workaround, which
with the latest update stopped working again?
https://github.com/libffi/libffi/issues/662
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=995223
Hi Adrian
,
John Paul Adrian Glaubitz wrote:
a) Upstream doesn't care and has not even looked at the issue yet:
That's worrying.
it'shttps://github.com/libffi/libffi/issues/662b) Debian's libffi maintainer hasn't understood the problem and assume
just a matter of passing the right configure option. It isn't so Ireopened
the bug reportÖ
I didn't notice it was reopened, but indeed, he writes "believes" issue
is closed, but facts are different :)
myself, I'mhttps://bugs.debian.org/cgi-bin/bugreport.cgi?bug=995223I don't know what else to do and honestly, I can't fix all these bugs
not even paid for doing that.
Adrian, I wasn't implying you should do it and I thanks for your work, I
just noted that things broke again. First part is diagnosing the issue
and hoping that maintainers and/or upstream work on it, which was not
the case.
What could be of use? forking libffi and attempting to patch all
configure warnings and generate a big patch and a pull-request?
Riccardo
a) Upstream doesn't care and has not even looked at the issue yet:
https://github.com/libffi/libffi/issues/662b) Debian's libffi maintainer hasn't understood the problem and assume it's
just a matter of passing the right configure option. It isn't so I reopened
the bug reportÖ
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=995223I don't know what else to do and honestly, I can't fix all these bugs myself, I'm
not even paid for doing that.
Adrian, I wasn't implying you should do it and I thanks for your work, I
just noted that things broke again. First part is diagnosing the issue
and hoping that maintainers and/or upstream work on it, which was not
the case.
What could be of use? forking libffi and attempting to patch all
configure warnings and generate a big patch and a pull-request?
[1] https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=ppc64&ver=3.4.2-3&stamp=1633957534&raw=0
b) Debian's libffi maintainer hasn't understood the problem and assume it's
just a matter of passing the right configure option. It isn't so I reopened
the bug reportÖ
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=995223
i got the 3.4.2-3+ports package after apt upgrade -t experimental and there are no illegal instructions
On 10/25/21 12:11, John Paul Adrian Glaubitz wrote:
b) Debian's libffi maintainer hasn't understood the problem and assumeit's
just a matter of passing the right configure option. It isn't so Ireopened
the bug reportÖ
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=995223
So, there is an update. A fellow Debian Developer suggested on
#debian-devel that
the configure option "--without-gcc-arch" will actually fix the problem
and it
did.
I will build the package manually with that option and upload it, the Debian's
libffi maintainer will hopefully shortly follow up with an updated package.
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org
`. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
It appears that when building portable code, the ax_gcc_archflag.m4 macro clears -mcpu for almost every arch except powerpc:
https://github.com/libffi/libffi/blob/master/m4/ax_gcc_archflag.m4#L241
case $host_cpu in i*86|x86_64*|amd64*) flag_prefixes="$flag_prefixes
-mcpu= -m";; esac
The fact that PowerPC is excluded must result in m4 passing the “-mcpu=“ flag matching the buildbot’s CPU, and breaks everything older than the buildbot.
This must happen with any build that uses this ax_gcc_archflag.m4 macro.
I wonder why powerpc* is excluded? Seems like this line should rather be:
case $host_cpu in i*86|x86_64*|powerpc*|amd64*) flag_prefixes="$flag_prefixes -mcpu= -m";; esac
Ken
PS. Although I did not as yet sort out getting the “experimental” packages
(I keep getting an error when I try to use that option), building libffi in
a couple of minutes on the local machine and installing that of course
works easily. — K
On Nov 2, 2021, at 2:01 PM, John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> wrote:
On 11/2/21 21:59, Cameron MacPherson wrote:
i got the 3.4.2-3+ports package after apt upgrade -t experimental and there
are no illegal instructions
As I expected. The build log didn't have any traces of "-mcpu=power8".
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org
`. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
On Nov 3, 2021, at 6:43 PM, Ken Cunningham <ken.cunningham.webuse@gmail.com> wrote:
It appears that when building portable code, the ax_gcc_archflag.m4 macro clears -mcpu for almost every arch except powerpc:
https://github.com/libffi/libffi/blob/master/m4/ax_gcc_archflag.m4#L241
case $host_cpu in i*86|x86_64*|amd64*) flag_prefixes="$flag_prefixes -mcpu= -m";; esac
The fact that PowerPC is excluded must result in m4 passing the “-mcpu=“ flag matching the buildbot’s CPU, and breaks everything older than the buildbot.
On Nov 2, 2021, at 2:01 PM, John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de> wrote:
On 11/2/21 21:59, Cameron MacPherson wrote:
i got the 3.4.2-3+ports package after apt upgrade -t experimental and there >> are no illegal instructions
As I expected. The build log didn't have any traces of "-mcpu=power8".
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org
`. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
On Nov 3, 2021, at 11:01 AM, Cameron MacPherson <cameron.macpherson@gmail.com> wrote:
you have to add the following to your /etc/apt/sources.list to get the experimental packages
deb http://ftp.ports.debian.org/debian-ports/ <http://ftp.ports.debian.org/debian-ports/> experimental main
see here
https://wiki.debian.org/DebianExperimental <https://wiki.debian.org/DebianExperimental>
On Wed, Nov 3, 2021, 10:42 AM Ken Cunningham <ken.cunningham.webuse@gmail.com <mailto:ken.cunningham.webuse@gmail.com>> wrote:
It appears that when building portable code, the ax_gcc_archflag.m4 macro clears -mcpu for almost every arch except powerpc:
https://github.com/libffi/libffi/blob/master/m4/ax_gcc_archflag.m4#L241 <https://github.com/libffi/libffi/blob/master/m4/ax_gcc_archflag.m4#L241>
case $host_cpu in i*86|x86_64*|amd64*) flag_prefixes="$flag_prefixes -mcpu= -m";; esac
The fact that PowerPC is excluded must result in m4 passing the “-mcpu=“ flag matching the buildbot’s CPU, and breaks everything older than the buildbot.
This must happen with any build that uses this ax_gcc_archflag.m4 macro.
I wonder why powerpc* is excluded? Seems like this line should rather be:
case $host_cpu in i*86|x86_64*|powerpc*|amd64*) flag_prefixes="$flag_prefixes -mcpu= -m";; esac
Ken
PS. Although I did not as yet sort out getting the “experimental” packages (I keep getting an error when I try to use that option), building libffi in a couple of minutes on the local machine and installing that of course works easily. — K
On Nov 2, 2021, at 2:01 PM, John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de <mailto:glaubitz@physik.fu-berlin.de>> wrote:
On 11/2/21 21:59, Cameron MacPherson wrote:
i got the 3.4.2-3+ports package after apt upgrade -t experimental and there
are no illegal instructions
As I expected. The build log didn't have any traces of "-mcpu=power8".
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org <mailto:glaubitz@debian.org> `. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de <mailto:glaubitz@physik.fu-berlin.de>
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
<br class=""></div><div class="">Best,</div><div class=""><br class=""></div><div class="">Ken<br class=""><div><br class=""><blockquote type="cite" class=""><div class="">On Nov 3, 2021, at 11:01 AM, Cameron MacPherson <<a href="mailto:cameron.macpherson@gmail.com" class="">cameron.macpherson@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="auto" class=""><div dir="auto" class="">you have to add the following to your /etc/apt/sources.list to get the
How is the current situation? I had an "fixed" package installed (I don't remember
if it was your build or one of mines) that got replaced during an upgrade because
it had a newver version and it is still broken.
http://ftp.ports.debian.org/debian-ports/pool-ppc64/main/libf/libffi/
The problem is not exclusive to PowerPC but also SPARC and 32-bit x86. It used to work before with version 3.4.2 in experimental:
See: https://buildd.debian.org/status/fetch.php?pkg=libffi&arch=powerpc&ver=3.4.2-1&stamp=1626443428&raw=0
Im pretty confident that my analysis is correct and the problem is triggered by the new autoconf version.
On 12/11/21 19:14, Riccardo Mottola wrote:
How is the current situation? I had an "fixed" package installed (I don't remember
if it was your build or one of mines) that got replaced during an upgrade because
it had a newver version and it is still broken.
There is a +ports package on ppc64:
http://ftp.ports.debian.org/debian-ports/pool-ppc64/main/libf/libffi/
Does this issue affect you on 32-bit powerpc as well? If yes, I need to manually update
the 32-bit package as well.
Matthias hasn't updated the libffi package yet, despite there being an RC bug open with
instructions on how to fix it. I have already pinged him several times.
I ran into the issue on my 32-bit powerpc:
[ 92.119397] unattended-upgr[627]: illegal instruction (4) at ed78a00 nip ed78a00 lr ed789f8 code 1 in libffi.so.8.1.0[ed76000+9000]
[ 92.119499] unattended-upgr[627]: code: 83e1001c 38210020 7c0803a6 4e800020 3d205858 38800000 61295858 387f01e8
[ 92.119516] unattended-upgr[627]: code: 913f0000 48004c3d 393f0014 7fa3eb78 <7c004eee> 393f01e4 7c004fae 48004b55
(ins)efraim@g4:~$ apt-cache policy libffi8
libffi8:
Installed: 3.4.2-3
Candidate: 3.4.2-3
Version table:
*** 3.4.2-3 500
500 http://deb.debian.org/debian-ports sid/main powerpc Packages
100 /var/lib/dpkg/status
I haven't dug into it at all yet.
I forgot to upload the manually built package for 32-bit powerpc which I have done
now. The fixed package should be available on the FTP servers within the next six
hours.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 295 |
Nodes: | 16 (2 / 14) |
Uptime: | 01:01:08 |
Calls: | 6,642 |
Calls today: | 2 |
Files: | 12,190 |
Messages: | 5,325,417 |