Opened 2 years ago
Closed 22 months ago
#65710 closed defect (fixed)
graphene @1.10.8_0+universal: graphene-1.0.pc differs and cannot be merged
| Reported by: | thetrial (alabay) | Owned by: | dbevans (David B. Evans) |
|---|---|---|---|
| Priority: | Normal | Milestone: | |
| Component: | ports | Version: | 2.7.2 |
| Keywords: | elcapitan legacy-os | Cc: | fhgwright (Fred Wright) |
| Port: | graphene |
Description
I have a strange error that seems to have do with architecture type.
:info:destroot fatal error: /Library/Developer/CommandLineTools/usr/bin/lipo: can't figure out the architecture type of: /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_graphene/graphene/work/destroot-i386//opt/local/lib/pkgconfig/graphene-1.0.pc
Something does not fit:
:info:destroot error: /Library/Developer/CommandLineTools/usr/bin/libtool: file: /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_graphene/graphene/work/destroot-i386//opt/local/lib/pkgconfig/graphene-1.0.pc is not an object file (not allowed in a library) :info:destroot error: /Library/Developer/CommandLineTools/usr/bin/libtool: file: /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_graphene/graphene/work/destroot-x86_64//opt/local/lib/pkgconfig/graphene-1.0.pc is not an object file (not allowed in a library)
and
:error:destroot Failed to destroot graphene: graphene-1.0.pc differs in /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_graphene/graphene/work/destroot-i386//opt/local/lib/pkgconfig and /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_graphene/graphene/work/destroot-x86_64//opt/local/lib/pkgconfig and cannot be merged :debug:destroot Error code: NONE
Logfile is attached.
Attachments (3)
Change History (15)
Changed 2 years ago by thetrial (alabay)
comment:1 Changed 2 years ago by ryandesign (Ryan Carsten Schmidt)
comment:2 Changed 2 years ago by thetrial (alabay)
Done. The names differ, but in real they are the same. And the diff is quite minimal … but interesting.
comment:3 Changed 2 years ago by fhgwright (Fred Wright)
Indeed this has the effect that graphene +universal can't be installed or upgraded. A second attempt without cleaning gets a different error, but still fails.
comment:4 Changed 2 years ago by fhgwright (Fred Wright)
| Cc: | fhgwright added |
|---|
comment:5 Changed 2 years ago by WaywardHeart
Here's the relevant commit in the graphene repo. Apparently SSE2 in i386 graphene was causing issues with gtk in linux. It looks like this issue occurred before in #56737, where it was decided to force SSE2 on for all Intel archs, but if SSE2 being problematic in i386 isn't just a linux thing then I guess it should be forced off for universal builds.
comment:6 Changed 2 years ago by fhgwright (Fred Wright)
Does the SSE2 option affect the API in an incompatible way? I.e., would claiming SSE2 support without actually having it cause problems (other than lower-than-expected performance)? If not, then maybe it could build the X86_64 slice with SSE2 and the i386 slice without, while claiming SSE2 support in the .pc file.
comment:8 Changed 22 months ago by thetrial (alabay)
Any idea, when graphene will be buildable again?! Abandoned?
comment:9 Changed 22 months ago by kencu (Ken)
SSE2 has been disabled for i386 builds for graphene by upstream, as mentioned above, apparently because on linux i386 there was not always 16 byte alignment for the variables, so there were segfaults.
Darwin/MacOS always guarantees at least 16byte alignment for malloc'd variables, so this cannot (should not) happen on Darwin.
For a number of years, since 2018, MacPorts built graphene with SSE2 support enabled in both the i386 and x86_64 builds. This happened because of the exact same issue with muniversal we are seeing in this ticket, and Marcus just disabled all the SSE2 special tests in graphene and let the compiler defines sort it out:
https://github.com/macports/macports-ports/commit/72c820b27ee817a55317ae20d8417a818a56af61
and this worked, it seemed, and everyone was happy.
(The test suite in graphene has never worked in MacPorts, so we could never prove it was OK, but nobody complained at least).
Recently Dave updated graphene and removed the disabler on this x86_64 test, so we are back at square one with the same exact problem building universal as before.
I would propose we again disable the x64_64 test in meson.build (10 seconds max, trivial to do, works to build universal) and let the __SSE2__ compiler define sort it out like it has all these past years for us.
That seems better to me that disabling the SIMD instruction set needlessly on these older CPUs that probably desperately need it, and we used the SIMD i386 instructions for years to support that idea.
comment:10 Changed 22 months ago by kencu (Ken)
here is one way to enable the tests in graphene to work:
% git diff
diff --git a/graphics/graphene/Portfile b/graphics/graphene/Portfile
index c0b220c329b..7112e88c7f4 100644
--- a/graphics/graphene/Portfile
+++ b/graphics/graphene/Portfile
@@ -48,6 +48,16 @@ configure.args-append \
-Dtests=false \
-Dinstalled_tests=false
+test.run yes
+test.target test
+pre-configure {
+ system -W ${worksrcpath} "${prefix}/bin/meson subprojects update --reset"
+
+}
+configure.args-replace -Dtests=false -Dtests=true \
+ -Dinstalled_tests=false -Dinstalled_tests=true
+
+
# the original motivation for this somewhat unusual pattern of supported_archs being set
# was here: https://github.com/macports/macports-ports/pull/2095#issuecomment-401559487
for a PR, that would have to wrapped up in the test variant block, of course. But with that, at least you can run the tests when building as i386...
comment:11 Changed 22 months ago by kencu (Ken)
comment:12 Changed 22 months ago by kencu (Ken)
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
Replying to thetrial:
Ignore those.
This is what's relevant. Please attach the two graphene-1.0.pc files so that we can see how they differ.