Bug 348909

Stretch targets are day 1 support for Firefox and the Swift runtime.

In addition to adding support for DARWIN_10_11 in configure.ac the following locations will definitely need to be looked at, as they don't appear to use a > platform approach (which would at least include OS X 10.11 in the set, if even incorrectly):

$ grep -r --exclude-dir=.svn "DARWIN_VERS == DARWIN_10_10" .
./coregrind/fixup_macho_loadcmds.c:#  if DARWIN_VERS == DARWIN_10_10
./coregrind/m_main.c:#if defined(VGO_darwin) && DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/priv_syswrap-darwin.h:#elif DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/priv_syswrap-darwin.h:#if DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/priv_syswrap-darwin.h:#endif /* DARWIN_VERS == DARWIN_10_10 */
./coregrind/m_syswrap/priv_syswrap-darwin.h:#if DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/priv_syswrap-darwin.h:#endif /* DARWIN_VERS == DARWIN_10_10 */
./coregrind/m_syswrap/syswrap-amd64-darwin.c:#  elif DARWIN_VERS == DARWIN_10_8 || DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/syswrap-amd64-darwin.c:#      elif DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/syswrap-darwin.c:#  if DARWIN_VERS == DARWIN_10_10 && VG_WORDSIZE == 8
./coregrind/m_syswrap/syswrap-darwin.c:#  endif /* DARWIN_VERS == DARWIN_10_10 && VG_WORDSIZE == 8 */
./coregrind/m_syswrap/syswrap-darwin.c:#     if DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/syswrap-x86-darwin.c:#  elif DARWIN_VERS == DARWIN_10_8 || DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./coregrind/m_syswrap/syswrap-x86-darwin.c:#     elif DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./include/vki/vki-darwin.h:#if DARWIN_VERS == DARWIN_10_10
./include/vki/vki-darwin.h:#endif /* DARWIN_VERS == DARWIN_10_10 */
./include/vki/vki-scnums-darwin.h:#elif DARWIN_VERS == DARWIN_10_10
./include/vki/vki-scnums-darwin.h:#elif DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./shared/vg_replace_strmem.c:# if DARWIN_VERS == DARWIN_10_9 || DARWIN_VERS == DARWIN_10_10
./VEX/priv/guest_amd64_toIR.c:#     if defined(VGP_amd64_darwin) && DARWIN_VERS == DARWIN_10_10

Created attachment 93161 [details]
Preliminary patchset to support OS X 10.11 El Capitan

== 592 tests, 229 stderr failures, 17 stdout failures, 0 stderrB failures, 0 stdoutB failures, 30 post failures ==

Which is relative to OS X 10.10:
stderr (+7)
stdout (+3)
post failures (unchanged)

Developer Preview 2, with Darwin kernel xnu-3247.1.6.2.2~1 reports the same, but importantly not worse, regressions relative to OS X 10.10 with the above preliminary patchset.

Rhys, thanks for looking at this.  I think that you should commit these
patches, if you feel they are suitable, so that others can easily try the
tree on 10.11.

I'll give the patches another visual review, then commit.

Above patchset landed in r15403.

And landed just in time. Apple have released a wider public beta of OS X 10.11, thus Valgrind SVN on that platform will see increasing adoption and testing.

Google currently top lists this bug report for most sensible search terms, for the inevitable bug reports that will start trickling in.

As of the final OS X 10.11 release and Valgrind 3.11.0 the regression test reports are:

== 596 tests, 224 stderr failures, 10 stdout failures, 0 stderrB failures, 0 stdoutB failures, 30 post failures ==

Which is relative to OS X 10.10 on the same hardware:
stderr (+6)
stdout (unchanged)
post failures (unchanged)

The majority of the failing stderr regression tests are within massif, helgrind and drd.

OS X 10.11 (El Capitan) does not allow user links in /usr/bin. It is cleared from non-Apple things and is further locked down using some new mechanism: I.e. not even root can put links there.

Valgrind plug-in for Eclipse should have configurable path to Valgrind installation or change the search location to /usr/local/bin

(In reply to Gorazd Hribar Rajteric from comment #10)
> OS X 10.11 (El Capitan) does not allow user links in /usr/bin. It is cleared
> from non-Apple things and is further locked down using some new mechanism:
> I.e. not even root can put links there.
> 
> Valgrind plug-in for Eclipse should have configurable path to Valgrind
> installation or change the search location to /usr/local/bin

I'm sure that's very useful information, but shouldn't you be telling whoever writes the valgrind plugin for Eclipse? There's not much we can do about it...

(In reply to Tom Hughes from comment #11)
> (In reply to Gorazd Hribar Rajteric from comment #10)
> > OS X 10.11 (El Capitan) does not allow user links in /usr/bin. It is cleared
> > from non-Apple things and is further locked down using some new mechanism:
> > I.e. not even root can put links there.
> > 
> > Valgrind plug-in for Eclipse should have configurable path to Valgrind
> > installation or change the search location to /usr/local/bin
> 
> I'm sure that's very useful information, but shouldn't you be telling
> whoever writes the valgrind plugin for Eclipse? There's not much we can do
> about it...

Tom, you're absolutely right. Bug 480757 (https://bugs.eclipse.org/bugs/show_bug.cgi?id=480757) has been created for Linux-Tools group.