David Chisnall
2015-Feb-18 20:41 UTC
[LLVMdev] [cfe-dev] [3.6 Release] RC3 has been tagged
On 18 Feb 2015, at 20:18, Jack Howarth <howarth.mailing.lists at gmail.com> wrote:> > Also doesn't FreeBSD use the standard major llvm releases for their > system compilers?We do pull in releases, but Dimitry also cherry-picks bug fixes throughout the life of each release. David
On 18 February 2015 at 20:41, David Chisnall <David.Chisnall at cl.cam.ac.uk> wrote:> We do pull in releases, but Dimitry also cherry-picks bug fixes throughout the life of each release.Do you guys recommend those same fixes for dot-releases? I'd like to track and include them. --renato
On Wed, Feb 18, 2015 at 3:41 PM, David Chisnall <David.Chisnall at cl.cam.ac.uk> wrote:> On 18 Feb 2015, at 20:18, Jack Howarth <howarth.mailing.lists at gmail.com> wrote: >> >> Also doesn't FreeBSD use the standard major llvm releases for their >> system compilers? > > We do pull in releases, but Dimitry also cherry-picks bug fixes throughout the life of each release.Looking at the Portfile's for the llvm releases at MacPort's, they are relying on the standard releases and haven't attempted to back port selective patches from upstream. Perhaps they need to start.> > David >
On 18 February 2015 at 21:03, Jack Howarth <howarth.mailing.lists at gmail.com> wrote:> Looking at the Portfile's for the llvm releases at MacPort's, they are > relying on the standard releases and haven't attempted to back port > selective patches from upstream. Perhaps they need to start.Or, perhaps, we should be giving more thought at the release process... Perhaps we should allow longer release cycles, overlap releases, etc. But this *will* incur in added complexity. It seems there is more interest in stable releases today than two years ago when we barely got enough "votes" to do dot-releases. If those new users care that much about a stable and non-regressing release, maybe we ought to get all of you guys on board in the next release. But that means you will have to actively contribute to fixing regressions and performance. All in all, the more testers and fixers, the better the release will be. But it's not fair to add complexity to the process if not enough people are willing to put an extra effort. cheers, --renato
On Wed, Feb 18, 2015 at 08:46:20PM +0000, Renato Golin wrote:> On 18 February 2015 at 20:41, David Chisnall > <David.Chisnall at cl.cam.ac.uk> wrote: > > We do pull in releases, but Dimitry also cherry-picks bug fixes throughout the life of each release. > > Do you guys recommend those same fixes for dot-releases? I'd like to > track and include them.All our patches to the base system llvm/clang are here: https://svnweb.freebsd.org/base/head/contrib/llvm/patches/ Most of the cherry-picks should likely be merged. We do have a different ABI stability standard so it may be that not all can be merged. For FreeBSD packages, I use release tarballs with as few patches as I can get away with. -- Brooks -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 181 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150218/513da9fe/attachment.sig>
On 18 February 2015 at 15:46, Renato Golin <renato.golin at linaro.org> wrote:> On 18 February 2015 at 20:41, David Chisnall > <David.Chisnall at cl.cam.ac.uk> wrote: >> We do pull in releases, but Dimitry also cherry-picks bug fixes throughout the life of each release. > > Do you guys recommend those same fixes for dot-releases? I'd like to > track and include them.In the past we did not, but Dimitry started requesting fixes with 3.5.1. As the LLVM release model has been evolving over time, so has our process for bringing Clang/LLVM/LLDB into FreeBSD. We used to carry around a large set of both local changes and backported fixes, but have been working to make that smaller and smaller as we move forward. (The 3.5.1 request is here, for interest: http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-November/079072.html)