Joel Jones via llvm-dev
2018-Jun-04 14:49 UTC
[llvm-dev] [LLD] Lack of REQUIRES causing test failures when not building with x86 support
If I build lld without x86 support (i.e. with "cmake -DLLVM_TARGETS_TO_BUILD=AArch64”) there are 117 tests that fail “make check-lld”. Upon examination, all are x86 related. There are multiple (884) tests that have “REQUIRES:.*x86” in the tests directory. Was there a cunning plan regarding the use of REQUIRES that I couldn’t discern? It seems that either 117 “REQUIRES” need to be added or 884 need to be removed. Joel Jones
Peter Smith via llvm-dev
2018-Jun-04 15:53 UTC
[llvm-dev] [LLD] Lack of REQUIRES causing test failures when not building with x86 support
Hello Joel, I don't think that there is a cunning plan with REQUIRES. I seem to remember that there was some debate over the testing strategy and platforms, but unfortunately I can't quickly find a link to it. In summary: - Much of the functionality of LLD is target independent, for example linker script support. - To write a test needs us to pick a target for these types of test and X86 was chosen as this was the most likely target to be compiled into build machines, for example the Arm and AArch64 buildbots include the X86 backend. I think that the missing 117 "REQUIRES" are an oversight, or historical artefact, and I think patches would be welcome to add them back in; a bot that doesn't build in the X86 backend is probably needed to keep this maintained as it is easy to forget the REQUIRES. However I think that if the goal is to test the majority of the functionality of LLD it would be wise to compile in support for X86 as much of the generic functionality will go untested. Peter On 4 June 2018 at 15:49, Joel Jones via llvm-dev <llvm-dev at lists.llvm.org> wrote:> If I build lld without x86 support (i.e. with "cmake -DLLVM_TARGETS_TO_BUILD=AArch64”) there are 117 tests that fail “make check-lld”. Upon examination, all are x86 related. > > There are multiple (884) tests that have “REQUIRES:.*x86” in the tests directory. Was there a cunning plan regarding the use of REQUIRES that I couldn’t discern? > > It seems that either 117 “REQUIRES” need to be added or 884 need to be removed. > > Joel Jones > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
via llvm-dev
2018-Jun-04 15:58 UTC
[llvm-dev] [LLD] Lack of REQUIRES causing test failures when not building with x86 support
> -----Original Message----- > From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Peter > Smith via llvm-dev > Sent: Monday, June 04, 2018 11:53 AM > To: Joel Jones > Cc: llvm-dev > Subject: Re: [llvm-dev] [LLD] Lack of REQUIRES causing test failures when > not building with x86 support > > Hello Joel, > > I don't think that there is a cunning plan with REQUIRES. I seem to > remember that there was some debate over the testing strategy and > platforms, but unfortunately I can't quickly find a link to it. In > summary: > - Much of the functionality of LLD is target independent, for example > linker script support. > - To write a test needs us to pick a target for these types of test > and X86 was chosen as this was the most likely target to be compiled > into build machines, for example the Arm and AArch64 buildbots include > the X86 backend. > > I think that the missing 117 "REQUIRES" are an oversight, or > historical artefact, and I think patches would be welcome to add them > back in; a bot that doesn't build in the X86 backend is probably > needed to keep this maintained as it is easy to forget the REQUIRES.A tactic used in some of the LLVM libraries is to create target-specific subdirectories. Tests that depend on a specific target go into the corresponding subdirectory. A small bit of lit code in the config file for the subdirectory will implicitly mark all tests UNSUPPORTED unless the correct target support is available. This avoids having to remember to add REQUIRES to the test, although you do still have to put it into the correct subdirectory. --paulr> However I think that if the goal is to test the majority of the > functionality of LLD it would be wise to compile in support for X86 as > much of the generic functionality will go untested. > > Peter > > On 4 June 2018 at 15:49, Joel Jones via llvm-dev > <llvm-dev at lists.llvm.org> wrote: > > If I build lld without x86 support (i.e. with "cmake - > DLLVM_TARGETS_TO_BUILD=AArch64”) there are 117 tests that fail “make > check-lld”. Upon examination, all are x86 related. > > > > There are multiple (884) tests that have “REQUIRES:.*x86” in the tests > directory. Was there a cunning plan regarding the use of REQUIRES that I > couldn’t discern? > > > > It seems that either 117 “REQUIRES” need to be added or 884 need to be > removed. > > > > Joel Jones > > _______________________________________________ > > LLVM Developers mailing list > > llvm-dev at lists.llvm.org > > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
Rui Ueyama via llvm-dev
2018-Jun-05 15:47 UTC
[llvm-dev] [LLD] Lack of REQUIRES causing test failures when not building with x86 support
In order to catch a missing "REQUIRES: x86", it might be a good idea to set up a build bot without the x86 support. On Mon, Jun 4, 2018 at 8:53 AM Peter Smith via llvm-dev < llvm-dev at lists.llvm.org> wrote:> Hello Joel, > > I don't think that there is a cunning plan with REQUIRES. I seem to > remember that there was some debate over the testing strategy and > platforms, but unfortunately I can't quickly find a link to it. In > summary: > - Much of the functionality of LLD is target independent, for example > linker script support. > - To write a test needs us to pick a target for these types of test > and X86 was chosen as this was the most likely target to be compiled > into build machines, for example the Arm and AArch64 buildbots include > the X86 backend. > > I think that the missing 117 "REQUIRES" are an oversight, or > historical artefact, and I think patches would be welcome to add them > back in; a bot that doesn't build in the X86 backend is probably > needed to keep this maintained as it is easy to forget the REQUIRES. > However I think that if the goal is to test the majority of the > functionality of LLD it would be wise to compile in support for X86 as > much of the generic functionality will go untested. > > Peter > > On 4 June 2018 at 15:49, Joel Jones via llvm-dev > <llvm-dev at lists.llvm.org> wrote: > > If I build lld without x86 support (i.e. with "cmake > -DLLVM_TARGETS_TO_BUILD=AArch64”) there are 117 tests that fail “make > check-lld”. Upon examination, all are x86 related. > > > > There are multiple (884) tests that have “REQUIRES:.*x86” in the tests > directory. Was there a cunning plan regarding the use of REQUIRES that I > couldn’t discern? > > > > It seems that either 117 “REQUIRES” need to be added or 884 need to be > removed. > > > > Joel Jones > > _______________________________________________ > > LLVM Developers mailing list > > llvm-dev at lists.llvm.org > > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20180605/488c932c/attachment.html>
Maybe Matching Threads
- [LLD] Lack of REQUIRES causing test failures when not building with x86 support
- [LLVMdev] [lld] [mach-o]: RFC: representing LC_REEXPORT_DYLIB
- samba 2.0.6 on HP-UX 11.0
- Rescuing svn repo from *IDIOTS*
- How to remove rows based on frequency of factor and then difference date scores