On 11/11/2020 23:26, Riccardo Mottola wrote:
It improves things definitely. Now gui apps start again.
Great!
Here the test-suite results:
76% tests passed, 44 tests failed out of 186
That's exactly the same number as before and the tests still fail, even
if in "real world" things improved? That's mysterious.
I thought we tested 32-bit in CI, but it turns out we test only 32-bit
Windows: for FreeBSD and Linux, we test only 64-bit x86. For another
project, we've set up some container thingies to use QEMU user mode for
testing other architectures - I'll see if I can set that up for libobjc2
so that we can test ARM as well. 32-bit builds ought to be easy anyway
just by adding -m32 to various flags.
I tried starting Ink and more complex GWorkspace, Gorm, GNUMail, they
all come up, load bundles, etc. So I think it is better than before and
branched.
Start 125: objc_msgSend
125/186 Test #125: objc_msgSend
......................................Child aborted***Exception: 0.00 sec
Start 126: objc_msgSend_optimised
126/186 Test #126: objc_msgSend_optimised
............................Child aborted***Exception: 0.00 sec
Start 127: objc_msgSend_legacy
127/186 Test #127: objc_msgSend_legacy
...............................Child aborted***Exception: 0.00 sec
Start 128: objc_msgSend_legacy_optimised
128/186 Test #128: objc_msgSend_legacy_optimised
.....................Child aborted***Exception: 0.00 sec
How can I give you more details on these (and the other) failures? there
are quite a lot.
Most of the tests run in a matrix of 4 configurations: debug / release
builds, v1 and v2 ABIs, so it looks like 11 real failures. I'll do a
32-bit build locally and take a look. It's a bit surprising
I don't know what differs from your setup, mine should be pretty
"standard" in my intentions.
ccmake . shows I am using clang:
CMAKE_CXX_COMPILER /usr/lib/llvm/10/bin/clang++
This should work fine, I probably have some bugs.
David