[haiku-commits] Re: r39308 - haiku/trunk/src/tests/system/runtime_loader/test_suite
- From: Ingo Weinhold <ingo_weinhold@xxxxxx>
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Fri, 05 Nov 2010 20:34:32 +0100
On 2010-11-05 at 19:06:28 [+0100], Rene Gollent <anevilyak@xxxxxxxxx> wrote:
> On Fri, Nov 5, 2010 at 1:53 PM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote:
> >
> > Just cd into the directory and run "./test_suite". Should take only a few
> > seconds.
>
> On 8.0:
Thanks for testing!
> load_resolve_basic1 ... ok
> load_lookup_next1 ... ok
> load_resolve_order1 ... ok
> load_resolve_order2 ... ok
> load_resolve_order3 ... ok
> load_resolve_order4 ... ok
> dlopen_resolve_basic1 ... ok
> dlopen_resolve_basic2 ... ok
> dlopen_resolve_basic3 ... ok
> dlopen_lookup_next1 ... ok
> dlopen_lookup_order1 ... ok
> dlopen_lookup_order2 ... test_run_ok: ./program: return value: 1, expected:
> 2
> FAILED
> dlopen_lookup_order3 ... ok
> dlopen_lookup_order4 ... ok
> dlopen_resolve_order1 ... Segmentation fault (core dumped)
> test_run_ok: ./program: return value: 139, expected: 1
> FAILED
> dlopen_resolve_order2 ... Segmentation fault (core dumped)
> test_run_ok: ./program: return value: 139, expected: 2
> FAILED
> dlopen_resolve_order3 ... ok
> dlopen_resolve_order4 ... Segmentation fault (core dumped)
> test_run_ok: ./program: return value: 139, expected: 2
> FAILED
> dlopen_resolve_order5 ... test_run_ok: ./program: return value: 1,
> expected: 2
> FAILED
> dlopen_resolve_order6 ... ok
> dlopen_resolve_order7 ... ok
I believe the failures of dlopen_lookup_order2 and dlopen_resolve_order5 are
somewhat expected, since FreeBSD seems to search loaded shared objects
differently -- breadth-first considering dlopen()ed libraries siblings of the
program, while Linux and Haiku search the dependency trees of the program and
each dlopen()ed library independently (also breadth-first).
The three crashing tests crash also crash on Linux.
On 2010-11-05 at 19:50:31 [+0100], scott mc <scottmc2@xxxxxxxxx> wrote:
> On Fri, Nov 5, 2010 at 10:06 AM, Rene Gollent <anevilyak@xxxxxxxxx> wrote:
> > On Fri, Nov 5, 2010 at 1:53 PM, Ingo Weinhold <ingo_weinhold@xxxxxx>
> > wrote:
> >>
> >> Just cd into the directory and run "./test_suite". Should take only a few
> >> seconds.
> >>
>
> On Haiku r39275 GCC2 build running in VMware:
Thanks! Haven't run them under Haiku myself yet.
> dlopen_resolve_order1, 2, and 4 each fail with cannot execute binary
> file, return value: 126, expected: 2
Might be the same issue as under Linux and FreeBSD, just that Haiku doesn't
support lazy symbol resolution and thus runs into the unresolvable symbol
already when loading the libraries and not when executing the code.
This is pretty weird, since I'm quite sure these tests worked on all three
systems two years ago.
CU, Ingo
Other related posts:
