Date: Sat, 08 Jul 2017 15:45:08 +0200 From: Matthew Rezny <rezny@freebsd.org> To: Jan Beich <jbeich@freebsd.org> Cc: ports-committers@freebsd.org, svn-ports-all@freebsd.org, svn-ports-head@freebsd.org Subject: Re: svn commit: r445016 - in head/x11/libxshmfence: . files Message-ID: <4854339.z0ub1xhMfD@workstation.reztek> In-Reply-To: <inj3-8577-wny@FreeBSD.org> References: <201707041347.v64Dlkx7028953@repo.freebsd.org> <2370475.W49dQvnFxO@workstation.reztek> <inj3-8577-wny@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday 08 July 2017 14:48:44 Jan Beich wrote: > Matthew Rezny <rezny@freebsd.org> writes: > >> ++#ifdef HAVE_MKOSTEMP > >> ++ fd = mkostemp(template, O_CLOEXEC); > >> ++#else > >> + fd = mkstemp(template); > >> ++#endif > >> + if (fd < 0) > >> + return fd; > >> + unlink(template); > > > > Please explain this unauthorized non-maintainer commit that lacks any PR > > or > > review. > > In PR 217676 the maintainer explicitly stated they didn't test DRI3 and > disregarded the issue. Maintainers are supposed to upstream changes > to prevent downstream-only bugs like we've seen here. > DRI3 is not supported by any release of FreeBSD so the support in ports is provided as a courtesy to those testing the drm-next effort but will not be fully supported until there is support in an OS release. There is a burden to supporting both and right now I must focus on ensuring everything works for DRI2, the common case, before DRI3 is even fully on much less fully tested. At the moment, that means coming up a proper fix for fallback to DRI2 in the EGL path since upstream has acknowledged the issue and requested a patch. > > Mentioning a closed PR does notyt count. This sort of change really > > needs to be reviewed by the maintainer(s) before it is committed. > > Understood. I'd ask you to not *modify* my patches without approval as well. > Let's not repeat beignet mistakes where upstream understood me better. Regarding beignet specifically, that is one I bumped along but cannot test. I've been trying to pick up everything under the x11 umbrella because it wasn't being done otherwise, but my interests (Xorg and Mesa, for Radeon, pre- GPU, and embedded/mobile hardware) are not as broad as the entire team needs to be; we could use more active members and/or offload some things that aren't really graphics, e.g. OpenCL. If you would like to take up Intel OpenCL, i.e. beignet, I think that would be beneficial to all. Trying to decipher the comments in various PRs without direct testing is certainly sub-optimal. > > Againb, I must assume this is wholly untested because there has been > > no mention of QA. > > QA is mandatory for committers. I did check DRI2 and DRI3 don't work > and don't orphan files. Which OS versions and on what platforms was it tested, and which of those were actually run tested vs merely build tested? This is information that should have been reported in the PR or review. Also, did you mean to say that you checked both DRI2 and DRI3 *do* work and don't orphan files?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4854339.z0ub1xhMfD>