Re: Can a leaf package require SSE2 on i386?
Jackson Doak writes (Re: Can a leaf package require SSE2 on i386?): The package infernal has also dropped i386 support for this reason. Using it's example, this can cause issues for downstreams with i386 arch:all builders. Just something to consider I think dropping the package from i386 is worse than all the other options, including the (very ugly indeed) situation where the package coredumps in its testsuite on some buildds. Any of the other options (debconf prompting in preinsts; proper error messages on program startup; disabling testsuites if sse2 not found; sse2-support feature `meta' package), singly or in combination, would be quite fine IMO. My advice would be to do what seems best for your package, and to continue to share information and experiences with other packages with similar problems. We can leave coming up with a uniform approach for another day. Ian. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/21529.36009.248682.262...@chiark.greenend.org.uk
Re: Can a leaf package require SSE2 on i386?
Le mercredi 17 septembre 2014 à 14:29 +0100, Ian Jackson a écrit : Jackson Doak writes (Re: Can a leaf package require SSE2 on i386?): The package infernal has also dropped i386 support for this reason. Using it's example, this can cause issues for downstreams with i386 arch:all builders. Just something to consider I think dropping the package from i386 is worse than all the other options, including the (very ugly indeed) situation where the package coredumps in its testsuite on some buildds. Any of the other options (debconf prompting in preinsts; proper error messages on program startup; disabling testsuites if sse2 not found; sse2-support feature `meta' package), singly or in combination, would be quite fine IMO. My advice would be to do what seems best for your package, and to continue to share information and experiences with other packages with similar problems. We can leave coming up with a uniform approach for another day. Given all the opinions expressed on the list (thanks!), I have decided to keep the i386 package and make it display an explicit error message at runtime if there is no SSE2 support in the processor. Concerning the buildd, I now realize that the package really needs SSE2 support at build time. The reason is that, Julia being a JIT-compiler, it is run at build time to create the binary image of its standard library. I'll see if that creates problems on the i386 buildds. -- .''`.Sébastien Villemot : :' :Debian Developer `. `' http://www.dynare.org/sebastien `- GPG Key: 4096R/381A7594 signature.asc Description: This is a digitally signed message part
Re: Can a leaf package require SSE2 on i386?
On 2014-09-17 15:47, Sébastien Villemot wrote: Concerning the buildd, I now realize that the package really needs SSE2 support at build time. The reason is that, Julia being a JIT-compiler, it is run at build time to create the binary image of its standard library. I'll see if that creates problems on the i386 buildds. buildds need to be 64bit-compatible anyway for multilib reasons, it will not cause problems, as all 64bit chips have SSE2 support. Kind regards Philipp Kern -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/19268b6c2d37e63e17486faef248c...@hub.kern.lc
Re: Subject=Re: Can a leaf package require SSE2 on i386?
On 15/09/2014 22:28, Barak A. Pearlmutter wrote: The package ikarus, another programming language implementation, also requires SSE2 support. There is a check in the preinst script which aborts installation if sse2 is unavailable. case $1 in install|upgrade) if egrep -q '^flags[[:space:]]*:.*\bsse2\b' /proc/cpuinfo; then # echo CPU instruction set extension sse2 confirmed true else echo error: CPU flag sse2 not found, aborting installation exit 1 fi ;; If this hack is really here, a bug should be filled. It is really a pain when something breaks when installing (or upgrading) a bunch of packages. This check should be moved at runtime. At install time, display a debconf notice if you really want to, but do not abruptly stop the installation without a really good reason (even when you try to remove the running kernel, perhaps leading to un unbootable system, you have the choice to abort or not) Regards, Vincent Cheers, --Barak. -- Vincent Danjean GPG key ID 0xD17897FA vdanj...@debian.org GPG key fingerprint: 621E 3509 654D D77C 43F5 CA4A F6AE F2AF D178 97FA Unofficial pkgs: http://moais.imag.fr/membres/vincent.danjean/deb.html APT repo: deb http://people.debian.org/~vdanjean/debian unstable main -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5417ddb1.1060...@free.fr
Re: Can a leaf package require SSE2 on i386?
On Tue, Sep 16, 2014 at 7:05 AM, Samuel Thibault sthiba...@debian.org wrote: Thomas Goirand, le Mon 15 Sep 2014 20:45:27 +0800, a écrit : On 09/15/2014 05:17 PM, Samuel Thibault wrote: Thomas Goirand, le Mon 15 Sep 2014 16:53:25 +0800, a écrit : I suppose (according to what's above) that using /usr/lib/sse4.2/x86_64-linux-gnu isn't supported (yet), right? I guess it shouldn't be hard to add the support, once the need is expressed :) Samuel Really? Ok... then *I NEED IT* ! :) As a bug report against libc6, I meant. Samuel -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140915230525.ga2...@type.youpi.perso.aquilenet.fr I meet the similar problem: a package (shine) set itself as mips32r2. I found it when I try dig out why it ftbfs on mips64el. -- YunQiang Su -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/cakcpw6xcheuauh44+przyhbfkau0xz10eg6zro9j1nkh98l...@mail.gmail.com
Re: Can a leaf package require SSE2 on i386?
On Sun, Sep 14, 2014 at 07:47:27AM +0100, Sébastien Villemot wrote: The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. A related topic was discussed during the bootstrap sprint[0], see section 2 (small part of that long mail). There, the topic of having optimized builds of packages was discussed. This technique can also be used to only provide optimized builds without providing non-optimized builds. It is not available in sid today. Please get in touch with Aron Xu (CCed). Helmut [0] https://lists.debian.org/20140829095214.gv19...@stoneboat.aleph1.co.uk -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140915075750.ga10...@alf.mars
Re: Can a leaf package require SSE2 on i386?
On Sun, Sep 14, 2014 at 8:47 AM, Sébastien Villemot sebast...@debian.org wrote: Hi, As the maintainer of julia (a technical computing language built on top of LLVM), I am wondering whether I should continue supporting the i386 architecture. The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. As per ld.so man page: HARDWARE CAPABILITIES Libraries might be compiled using hardware-specific instructions which do not exist on all CPU. Such libraries should be installed in directories whose name defines the hardware capabilities such as /usr/lib/sse2/. The dynamic linker checks these directories against the hardware of the machine and selects the best suitable ver‐ sion of a given library. Hardware capabilities directories could be cascaded to combine CPU features. Hardware capabilities depends on the CPU. The following names are currently recognized: [...] x86 (32-bit only) acpi, apic, clflush, cmov, cx8, dts, fxsr, ht, i386, i486, i586, i686, mca, mmx, mtrr, pat, pbe, pge, pn, pse36, sep, ss, sse, sse2, tm HTH -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CA+7wUszjeCz0hVVAQb9=evysce2q-axdqn3xvxlewdx84as...@mail.gmail.com
Re: Can a leaf package require SSE2 on i386?
On 09/15/2014 04:04 PM, Mathieu Malaterre wrote: On Sun, Sep 14, 2014 at 8:47 AM, Sébastien Villemot sebast...@debian.org wrote: Hi, As the maintainer of julia (a technical computing language built on top of LLVM), I am wondering whether I should continue supporting the i386 architecture. The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. As per ld.so man page: HARDWARE CAPABILITIES Libraries might be compiled using hardware-specific instructions which do not exist on all CPU. Such libraries should be installed in directories whose name defines the hardware capabilities such as /usr/lib/sse2/. The dynamic linker checks these directories against the hardware of the machine and selects the best suitable ver‐ sion of a given library. Hardware capabilities directories could be cascaded to combine CPU features. Hardware capabilities depends on the CPU. The following names are currently recognized: [...] x86 (32-bit only) acpi, apic, clflush, cmov, cx8, dts, fxsr, ht, i386, i486, i586, i686, mca, mmx, mtrr, pat, pbe, pge, pn, pse36, sep, ss, sse, sse2, tm HTH Thanks. That's nice, however, what if a library needs SSE4.2? Two of my packages, eg libjerasure2 and libgf-complete1, would benefit from major speed-up if there was SSE4 available. I have currently disabled all SSE stuff, and would really like to provide an SSE4.2 version as well. I suppose (according to what's above) that using /usr/lib/sse4.2/x86_64-linux-gnu isn't supported (yet), right? (note: I already asked upstream if it was possible to do runtime detection, and the answer is currently no, unfortunately) Cheers, Thomas Goirand (zigo) -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5416a905.3000...@debian.org
Re: Can a leaf package require SSE2 on i386?
Thomas Goirand, le Mon 15 Sep 2014 16:53:25 +0800, a écrit : I suppose (according to what's above) that using /usr/lib/sse4.2/x86_64-linux-gnu isn't supported (yet), right? I guess it shouldn't be hard to add the support, once the need is expressed :) Samuel -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140915091748.ge3...@type.bordeaux.inria.fr
Re: Can a leaf package require SSE2 on i386?
On 09/15/2014 05:17 PM, Samuel Thibault wrote: Thomas Goirand, le Mon 15 Sep 2014 16:53:25 +0800, a écrit : I suppose (according to what's above) that using /usr/lib/sse4.2/x86_64-linux-gnu isn't supported (yet), right? I guess it shouldn't be hard to add the support, once the need is expressed :) Samuel Really? Ok... then *I NEED IT* ! :) Thomas -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5416df67.9000...@debian.org
Re: Can a leaf package require SSE2 on i386?
On Sun, Sep 14, 2014 at 04:40:14PM +0200, Adrien Clerc wrote: I'm in the category of people who installed their Debian 8 years ago, on an old AMD processor, only i686. My hardware was upgraded since, but the system remains. I've searched for cross-grade, but nothing serious comes out, except reinstall everything. If you have some clear documentation (at least some steps, I'm curious enough to read some apt manuals), I think you can go on dropping i386 architecture. It can be done if you are competent and confident. I successfully crossgraded a production system from i386 to amd64 a while back, roughly following the steps from this blog post: http://blog.zugschlus.de/archives/972-How-to-amd64-an-i386-Debian-installation-with-multiarch.html The system in question was originally installed in 1999 (with potato, I think) and has been contiuously upgraded since then. Caveats from my experience, though: * I was doing this on a server, not a desktop. * I put some effort into removing obsolete or unnecessary packages first and making sure that amd64 versions of all my non-Debian packages were available, in order to simplify things. This procedure is much easier to execute on a system whose package database is nice and clean. * I started out by cloning the relevant parts of the system into a chroot, disabling daemon startup, and upgrading that, until I'd got far enough that I was comfortable I could finish the job. * I did this on a stable release. Doing this in testing or unstable where it's more likely that the archive might introduce new versions of Multi-Arch: same packages part-way through the job would have been rather more exciting. * I know dpkg and apt very well, was a contributor to the multiarch design, and am comfortable recovering things by hand if necessary. -- Colin Watson [cjwat...@debian.org] -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140915165025.ga12...@riva.ucam.org
Re: Can a leaf package require SSE2 on i386?
On 15. sep. 2014 18:50, Colin Watson wrote: On Sun, Sep 14, 2014 at 04:40:14PM +0200, Adrien Clerc wrote: I'm in the category of people who installed their Debian 8 years ago, on an old AMD processor, only i686. My hardware was upgraded since, but the system remains. I've searched for cross-grade, but nothing serious comes out, except reinstall everything. If you have some clear documentation (at least some steps, I'm curious enough to read some apt manuals), I think you can go on dropping i386 architecture. It can be done if you are competent and confident. I successfully crossgraded a production system from i386 to amd64 a while back, roughly following the steps from this blog post: Lacking competence, dumb luck and stable power for ~48 hrs will also work :~ . For production systems, cross-grade is still a bad idea, if only because you MUST plan for extended down-time, Cross-grade can be made to work on non-mission-critical boxes. This message is proof of that. :-. dpkg has cross-grade working. apt can be pressed into use by apt-get install --reinstall . Of course, if you do not use the appropriate tools for the job, you MUST have sufficient sed, awk and perl -fu. I had grep failing for a while, which made some .debs uninstallable. A large cache of downloaded .debs helps. I was also happy that I remembered seeing some android phones using 8.8.8.8 for recursive DNS, I never paid attention to what my ISP gives out. Then network at home (apart from the phones) was out for a couple of hours until I got bind working again. My system got a good rinse from it. Now there are hardly any packages from squeeze left, and the last SuSE 6.4 binaries got the boot :) (Now THAT was a truly painful transition, back in the day). This guide http://www.ewan.cc/?q=node/90 looks quite safe. If I had followed it I guess the pain would have been less than my route. Minimum requirement for reducing hassle is to make sure your existing i386 packages are the same version as the amd64 you will be replacing them with. Barring that you need to forcibly purge the old and insert then new, while keeping a running system. I also remembered about postgresql being fussy about bit-width too late, so I had to install a 32-bit chroot to get a fresh export from the entire install. Having a dump of the databases is well and good, but you ned the roles aswell. postgres-32-bit will NOT run on 64-bit libraries. RRD-databases are the same, so having a chroot ready for stuff like that is probably wise no matter how well you plan things. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/54172fa0.50...@alstadheim.priv.no
Subject=Re: Can a leaf package require SSE2 on i386?
The package ikarus, another programming language implementation, also requires SSE2 support. There is a check in the preinst script which aborts installation if sse2 is unavailable. case $1 in install|upgrade) if egrep -q '^flags[[:space:]]*:.*\bsse2\b' /proc/cpuinfo; then # echo CPU instruction set extension sse2 confirmed true else echo error: CPU flag sse2 not found, aborting installation exit 1 fi ;; Cheers, --Barak. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CANa01B+3oeZjFtTm=-t+herqb29ag-4uvfu3j8henfxto1h...@mail.gmail.com
Re: Can a leaf package require SSE2 on i386?
Thomas Goirand, le Mon 15 Sep 2014 20:45:27 +0800, a écrit : On 09/15/2014 05:17 PM, Samuel Thibault wrote: Thomas Goirand, le Mon 15 Sep 2014 16:53:25 +0800, a écrit : I suppose (according to what's above) that using /usr/lib/sse4.2/x86_64-linux-gnu isn't supported (yet), right? I guess it shouldn't be hard to add the support, once the need is expressed :) Samuel Really? Ok... then *I NEED IT* ! :) As a bug report against libc6, I meant. Samuel -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140915230525.ga2...@type.youpi.perso.aquilenet.fr
Can a leaf package require SSE2 on i386?
Hi, As the maintainer of julia (a technical computing language built on top of LLVM), I am wondering whether I should continue supporting the i386 architecture. The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. So I have two options: either ship a i386 package that only works on SSE2 processors (ideally giving a meaningful error message when run on older CPUs); or drop support for i386, which is a disservice to our users (the few who have a SSE2-capable but not x86-64-capable processor will be left out; and those who are running the i386 arch on a x86-64-capable processor will have to cross-grade to amd64 or at least use multi-arch with a 64-bit kernel). Also note that my understanding is that some i386 buildds are not SSE2-capable (because they are qemu guests configured as such). So, if I were to ship an i386 package requiring SSE2, the testsuite would fail on those buildds (meaning that I would have to make the testsuite non fatal, or ask for blacklisting of those buildd). What's your opinion on this issue? [1] https://github.com/JuliaLang/julia/issues/7185 -- .''`.Sébastien Villemot : :' :Debian Developer `. `' http://www.dynare.org/sebastien `- GPG Key: 4096R/381A7594 signature.asc Description: This is a digitally signed message part
Re: Can a leaf package require SSE2 on i386?
On Sun, Sep 14, 2014 at 2:47 AM, Sébastien Villemot wrote: The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. chromium upstream decided to go SSE2-only, but I've reverted that in the Debian packages for now. I would prefer to not diverge, and would do so if there were a convenient way to detect and prompt users about the problem (rather than segfault). Best wishes, Mike -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CANTw=MPZr2suB3DBN7jvADJpG2CWz8NECtnc_epe4hoe...@mail.gmail.com
Re: Can a leaf package require SSE2 on i386?
Hi, On Sep 14, 2014 9:16 AM, Michael Gilbert mgilb...@debian.org wrote: On Sun, Sep 14, 2014 at 2:47 AM, Sébastien Villemot wrote: The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. chromium upstream decided to go SSE2-only, but I've reverted that in the Debian packages for now. I would prefer to not diverge, and would do so if there were a convenient way to detect and prompt users about the problem (rather than segfault). How about creating a package named like sse2-support for i386 which fails to install (unless it is forced) on not SSE2-capable hardware emitting a proper error message? Packages requiring SSE2 could (build-) depend on it. BTW steam already requires SSE2. Cheers, Balint
Re: Can a leaf package require SSE2 on i386?
The package infernal has also dropped i386 support for this reason. Using it's example, this can cause issues for downstreams with i386 arch:all builders. Just something to consider On Sun, Sep 14, 2014 at 6:07 PM, Balint Reczey rbal...@gmail.com wrote: Hi, On Sep 14, 2014 9:16 AM, Michael Gilbert mgilb...@debian.org wrote: On Sun, Sep 14, 2014 at 2:47 AM, Sébastien Villemot wrote: The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. chromium upstream decided to go SSE2-only, but I've reverted that in the Debian packages for now. I would prefer to not diverge, and would do so if there were a convenient way to detect and prompt users about the problem (rather than segfault). How about creating a package named like sse2-support for i386 which fails to install (unless it is forced) on not SSE2-capable hardware emitting a proper error message? Packages requiring SSE2 could (build-) depend on it. BTW steam already requires SSE2. Cheers, Balint
Re: Can a leaf package require SSE2 on i386?
On Sun, Sep 14, 2014 at 8:47 AM, Sébastien Villemot sebast...@debian.org wrote: Hi, As the maintainer of julia (a technical computing language built on top of LLVM), I am wondering whether I should continue supporting the i386 architecture. The bottom line is that julia needs SSE2 (and porting it to the x87 FPU requires changes that are beyond what I am willing/able to do, see [1] for more details). And the presence of SSE2 is not guaranteed on the i386 architecture. So I have two options: either ship a i386 package that only works on SSE2 processors (ideally giving a meaningful error message when run on older CPUs); or drop support for i386, which is a disservice to our users (the few who have a SSE2-capable but not x86-64-capable processor will be left out; and those who are running the i386 arch on a x86-64-capable processor will have to cross-grade to amd64 or at least use multi-arch with a 64-bit kernel). Also note that my understanding is that some i386 buildds are not SSE2-capable (because they are qemu guests configured as such). So, if I were to ship an i386 package requiring SSE2, the testsuite would fail on those buildds (meaning that I would have to make the testsuite non fatal, or ask for blacklisting of those buildd). What's your opinion on this issue? Add sse/sse2 support to libmmx ? https://www-sop.inria.fr/members/Sylvain.Pion/progs/mmx-emu/ [1] https://github.com/JuliaLang/julia/issues/7185 -- .''`.Sébastien Villemot : :' :Debian Developer `. `' http://www.dynare.org/sebastien `- GPG Key: 4096R/381A7594 -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/cae2spabg3bqjkx9sosnz7b_687yygi837ftxuxmfkgy-ylq...@mail.gmail.com
Re: Can a leaf package require SSE2 on i386?
Le 14/09/2014 08:47, Sébastien Villemot a écrit : So I have two options: either ship a i386 package that only works on SSE2 processors (ideally giving a meaningful error message when run on older CPUs); or drop support for i386, which is a disservice to our users (the few who have a SSE2-capable but not x86-64-capable processor will be left out; and those who are running the i386 arch on a x86-64-capable processor will have to cross-grade to amd64 or at least use multi-arch with a 64-bit kernel). I'm in the category of people who installed their Debian 8 years ago, on an old AMD processor, only i686. My hardware was upgraded since, but the system remains. I've searched for cross-grade, but nothing serious comes out, except reinstall everything. If you have some clear documentation (at least some steps, I'm curious enough to read some apt manuals), I think you can go on dropping i386 architecture. Otherwise, even if I'm not a julia user, I'll be really sad to see a package lost only because SSE2 is needed. Adrien -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5415a8ce.1020...@antipoul.fr
Re: Can a leaf package require SSE2 on i386?
Le dimanche 14 septembre 2014 à 07:47 +0100, Sébastien Villemot a écrit : So I have two options: either ship a i386 package that only works on SSE2 processors (ideally giving a meaningful error message when run on older CPUs); or drop support for i386, which is a disservice to our users The silliness of the second option should be enough to make the choice. Mentioning in the description, and in an error message at launch time, that the package doesn’t work without SSE2, should be more than enough. Also note that my understanding is that some i386 buildds are not SSE2-capable (because they are qemu guests configured as such). So, if I were to ship an i386 package requiring SSE2, the testsuite would fail on those buildds (meaning that I would have to make the testsuite non fatal, or ask for blacklisting of those buildd). I have doubts about the existence of such qemu guests that wouldn’t be KVM-enabled. In any case, you have the option to disable the testsuite depending on the contents of /proc/cpuinfo. Cheers, -- .''`. Josselin Mouette : :' : `. `' `- -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/1410712506.10598.3.ca...@kagura.malsain.org