Tom Stellard via llvm-dev
2018-Dec-04 18:43 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
Hi, Fixing http://llvm.org/PR39427 in the release_70 branch, will change the ABI of a clang built libLLVM-7.so so that it is no longer compatible with the 7.0.0 release. libLLVM-7.so built by gcc will not be affected by this fix. Changing the ABI is something we aren't supposed to do in stable releases, but this fixes an ABI difference between clang and gcc built libLLVM-7.so that is impacting many users, including Linux distributions, so an ABI change for clang only seems like the least bad solution in this case. I am proposing that we hold this fix out of 7.0.1 and instead ship it in a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to are still able to get all the bug-fixes without the ABI changes. How does this sound to everyone? Thanks, Tom
Hans Wennborg via llvm-dev
2018-Dec-05 08:37 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
On Tue, Dec 4, 2018 at 7:43 PM Tom Stellard <tstellar at redhat.com> wrote:> > Hi, > > Fixing http://llvm.org/PR39427 in the release_70 branch, will change > the ABI of a clang built libLLVM-7.so so that it is no longer compatible > with the 7.0.0 release. libLLVM-7.so built by gcc will not be affected > by this fix. > > Changing the ABI is something we aren't supposed to do in stable releases, > but this fixes an ABI difference between clang and gcc built > libLLVM-7.so that is impacting many users, including Linux distributions, > so an ABI change for clang only seems like the least bad solution in this case. > > I am proposing that we hold this fix out of 7.0.1 and instead ship it in > a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to > are still able to get all the bug-fixes without the ABI changes. > > How does this sound to everyone?I think this sounds reasonable. The important thing is that the fix gets shipped, and if we do e.g. both 7.0.1 and 7.0.2 (I'm not sure what version numbers would be best to use) that should hopefully serve everyone well. Thanks, Hans
Jacob Lifshay via llvm-dev
2018-Dec-05 10:06 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
On Tue, Dec 4, 2018, 10:43 Tom Stellard via llvm-dev < llvm-dev at lists.llvm.org wrote:> I am proposing that we hold this fix out of 7.0.1 and instead ship it in > a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to > are still able to get all the bug-fixes without the ABI changes. >I think we should go with 7.1.0 instead of 7.0.2, since 7.1.0 looks unusual as opposed to 7.0.2, which looks like business as usual. This will help people to notice that the ABI changed.> > How does this sound to everyone? >Sounds reasonable to me. Jacob Lifshay>-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181205/0963a614/attachment.html>
Finkel, Hal J. via llvm-dev
2018-Dec-05 14:22 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
On 12/5/18 4:06 AM, Jacob Lifshay via llvm-dev wrote: On Tue, Dec 4, 2018, 10:43 Tom Stellard via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org> wrote: I am proposing that we hold this fix out of 7.0.1 and instead ship it in a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to are still able to get all the bug-fixes without the ABI changes. I think we should go with 7.1.0 instead of 7.0.2, since 7.1.0 looks unusual as opposed to 7.0.2, which looks like business as usual. This will help people to notice that the ABI changed. +1 - ABI changes should increment the minor version number. -Hal How does this sound to everyone? Sounds reasonable to me. Jacob Lifshay _______________________________________________ LLVM Developers mailing list llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev -- Hal Finkel Lead, Compiler Technology and Programming Languages Leadership Computing Facility Argonne National Laboratory -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181205/dac48407/attachment.html>
Kristina Brooks via llvm-dev
2018-Dec-05 14:54 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
Hi, While on topic of ABI changes, could I request that this (only the first part which bumped the constant number up for OS targets) be merged into an ABI breaking release? https://reviews.llvm.org/rL347832 This is only the LLVMSupport part of the patch, the Clang half depends on it but that is more of a feature thing so it's better left for 8.x.x. I'm mentioning this because there was some confusion w/r Debian merging this patch into their 7.x.x fork before it went into trunk which caused ABI breakage for anyone who relied on the last enum member to determine the size. Thank you. - Kristina Hans Wennborg via llvm-dev wrote:> On Tue, Dec 4, 2018 at 7:43 PM Tom Stellard <tstellar at redhat.com> wrote: >> >> Hi, >> >> Fixing http://llvm.org/PR39427 in the release_70 branch, will change >> the ABI of a clang built libLLVM-7.so so that it is no longer compatible >> with the 7.0.0 release. libLLVM-7.so built by gcc will not be affected >> by this fix. >> >> Changing the ABI is something we aren't supposed to do in stable releases, >> but this fixes an ABI difference between clang and gcc built >> libLLVM-7.so that is impacting many users, including Linux distributions, >> so an ABI change for clang only seems like the least bad solution in this case. >> >> I am proposing that we hold this fix out of 7.0.1 and instead ship it in >> a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to >> are still able to get all the bug-fixes without the ABI changes. >> >> How does this sound to everyone? > > I think this sounds reasonable. The important thing is that the fix > gets shipped, and if we do e.g. both 7.0.1 and 7.0.2 (I'm not sure > what version numbers would be best to use) that should hopefully serve > everyone well. > > Thanks, > Hans > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 3992 bytes Desc: S/MIME Cryptographic Signature URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181205/f0ec639b/attachment.bin>
Tom Stellard via llvm-dev
2018-Dec-05 19:39 UTC
[llvm-dev] ABI change in LLVM 7.0.x release
On 12/05/2018 02:06 AM, Jacob Lifshay wrote:> On Tue, Dec 4, 2018, 10:43 Tom Stellard via llvm-dev <llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org> wrote: > > I am proposing that we hold this fix out of 7.0.1 and instead ship it in > a special 7.0.2 (or maybe 7.1.0 ?) release, so that users that want to > are still able to get all the bug-fixes without the ABI changes. > > I think we should go with 7.1.0 instead of 7.0.2, since 7.1.0 looks unusual as opposed to 7.0.2, which looks like business as usual. This will help people to notice that the ABI changed. >This was my initial thought as well, but I'm not sure if this is technically possible with our current system. The shared object only contains the major version in its name, so we would need some other changes. I can look into how feasible this would be. -Tom> > How does this sound to everyone? > > Sounds reasonable to me. > > Jacob Lifshay >
Possibly Parallel Threads
- [Release-testers] LLVM 7.1.0 release - Please test the branch
- LLVM 7.1.0 release - Please test the branch
- [Release-testers] LLVM 7.1.0 release - Please test the branch
- [Release-testers] LLVM 7.1.0 release - Please test the branch
- [Release-testers] LLVM 7.1.0 release - Please test the branch