Message ID | 20180608225932.13127-7-joseph.kogut@gmail.com |
---|---|
State | Superseded |
Headers | show |
Series | chromium: new package | expand |
Joseph, all, On 09/06/2018 00:59, Joseph Kogut wrote: > Signed-off-by: Joseph Kogut <joseph.kogut@gmail.com> > --- > ...type-in-ORC-readMem-client-interface.patch | 37 ------------------- > package/llvm/llvm.hash | 2 +- > package/llvm/llvm.mk | 2 +- > 3 files changed, 2 insertions(+), 39 deletions(-) > delete mode 100644 package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch > > -LLVM_VERSION = 5.0.2 > +LLVM_VERSION = 6.0.0 > LLVM_SITE = http://llvm.org/releases/$(LLVM_VERSION) > LLVM_SOURCE = llvm-$(LLVM_VERSION).src.tar.xz > LLVM_LICENSE = NCSA Tested-by: Valentin Korenblit <valentin.korenblit@smile.fr> I've applied the patch and tested llvmpipe on ARM and x86. My only doubt is whether we should also keep llvm 5 (no longer mantained). Apparently there haven't been changes in the API, most changes are related to backends: https://releases.llvm.org/6.0.0/docs/ReleaseNotes.html. I ask this question because this topic was discussed during the meeting: https://elinux.org/Buildroot:DeveloperDaysFOSDEM2018#LLVM.2FClang I believe we could just bump to 6.0.0 as currently there is only Mesa using this package. Best regards, Valentin
Valentin, On Tue, Jun 12, 2018 at 7:34 AM Valentin Korenblit <valentin.korenblit@smile.fr> wrote: > > Joseph, all, > > On 09/06/2018 00:59, Joseph Kogut wrote: > > Signed-off-by: Joseph Kogut <joseph.kogut@gmail.com> > > --- > > ...type-in-ORC-readMem-client-interface.patch | 37 ------------------- > > package/llvm/llvm.hash | 2 +- > > package/llvm/llvm.mk | 2 +- > > 3 files changed, 2 insertions(+), 39 deletions(-) > > delete mode 100644 package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch > > > > -LLVM_VERSION = 5.0.2 > > +LLVM_VERSION = 6.0.0 > > LLVM_SITE = http://llvm.org/releases/$(LLVM_VERSION) > > LLVM_SOURCE = llvm-$(LLVM_VERSION).src.tar.xz > > LLVM_LICENSE = NCSA > > Tested-by: Valentin Korenblit <valentin.korenblit@smile.fr> > > I've applied the patch and tested llvmpipe on ARM and x86. My only > doubt is whether we should also keep llvm 5 (no longer mantained). > Apparently there haven't been changes in the API, most changes are > related to backends: https://releases.llvm.org/6.0.0/docs/ReleaseNotes.html. > > I ask this question because this topic was discussed during the meeting: > https://elinux.org/Buildroot:DeveloperDaysFOSDEM2018#LLVM.2FClang > I was curious about that as well, as I read the notes about the meeting. If LLVM 5.0 is no longer maintained, I don't see any reason to continue supporting it, especially if the API hasn't broken. > I believe we could just bump to 6.0.0 as currently there is only Mesa > using this package. > On the other hand, there's no way for us to know if someone is using LLVM in their own application that would break by upgrading. I think the answer to that problem is that they should stay on a Buildroot release branch that packages the version they need, or they should submit patches to allow the older version to be selected. With 6.0.0 being the current stable release, I do think it makes sense for us to bump it. Best, Joseph
diff --git a/package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch b/package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch deleted file mode 100644 index 0343b4c7e8..0000000000 --- a/package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch +++ /dev/null @@ -1,37 +0,0 @@ -From 72ea6ea635d5b5a88f411710daf7e1d340d232d8 Mon Sep 17 00:00:00 2001 -From: Tilmann Scheller <tschelle@redhat.com> -Date: Thu, 1 Feb 2018 11:40:01 -0600 -Subject: [PATCH] Fix return type in ORC readMem() client interface. - -GCC 8.0.1 detects the type mismatch and causes the compilation to fail. Clang -and earlier versions of GCC don't detect the issue. - -Fixes rhbz#1540620. - -This patch was taken from llvm5.0-5.0.1-7.fc28.src.rpm - -Link to bug: https://bugzilla.redhat.com/show_bug.cgi?id=1540620 - -Signed-off-by: Valentin Korenblit <valentin.korenblit@smile.fr> ---- - include/llvm/ExecutionEngine/Orc/OrcRemoteTargetClient.h | 4 ++-- - 1 file changed, 2 insertions(+), 2 deletions(-) - -diff --git a/include/llvm/ExecutionEngine/Orc/OrcRemoteTargetClient.h b/include/llvm/ExecutionEngine/Orc/OrcRemoteTargetClient.h -index da02250ba16..bed472e2e0e 100644 ---- a/include/llvm/ExecutionEngine/Orc/OrcRemoteTargetClient.h -+++ b/include/llvm/ExecutionEngine/Orc/OrcRemoteTargetClient.h -@@ -713,8 +713,8 @@ private: - - uint32_t getTrampolineSize() const { return RemoteTrampolineSize; } - -- Expected<std::vector<char>> readMem(char *Dst, JITTargetAddress Src, -- uint64_t Size) { -+ Expected<std::vector<uint8_t>> readMem(char *Dst, JITTargetAddress Src, -+ uint64_t Size) { - // Check for an 'out-of-band' error, e.g. from an MM destructor. - if (ExistingError) - return std::move(ExistingError); --- -2.14.3 - diff --git a/package/llvm/llvm.hash b/package/llvm/llvm.hash index 14f8bb4488..93bdf2e20e 100644 --- a/package/llvm/llvm.hash +++ b/package/llvm/llvm.hash @@ -1,3 +1,3 @@ # locally calculated -sha256 d522eda97835a9c75f0b88ddc81437e5edbb87dc2740686cb8647763855c2b3c llvm-5.0.2.src.tar.xz +sha256 1ff53c915b4e761ef400b803f07261ade637b0c269d99569f18040f3dcee4408 llvm-6.0.0.src.tar.xz sha256 abd4d8794808bacb1eb6924d49efafd9ab6eef88faaaeb5d3cfa13ee3670d672 LICENSE.TXT diff --git a/package/llvm/llvm.mk b/package/llvm/llvm.mk index 06f95c96f8..d2db05d94c 100644 --- a/package/llvm/llvm.mk +++ b/package/llvm/llvm.mk @@ -4,7 +4,7 @@ # ################################################################################ -LLVM_VERSION = 5.0.2 +LLVM_VERSION = 6.0.0 LLVM_SITE = http://llvm.org/releases/$(LLVM_VERSION) LLVM_SOURCE = llvm-$(LLVM_VERSION).src.tar.xz LLVM_LICENSE = NCSA
Signed-off-by: Joseph Kogut <joseph.kogut@gmail.com> --- ...type-in-ORC-readMem-client-interface.patch | 37 ------------------- package/llvm/llvm.hash | 2 +- package/llvm/llvm.mk | 2 +- 3 files changed, 2 insertions(+), 39 deletions(-) delete mode 100644 package/llvm/0001-Fix-return-type-in-ORC-readMem-client-interface.patch