Probably I've understood the reason of the segmentation fault on Linux platform.
I almost sure it is related with the build system. An argument why I think so is the following. The basic_istringstream() class from the /usr/include/g++/sstream include should be used to construct the configuration environment for the execution manager (vm/em/src/DrlEMIml.cpp file). However the recent build uses this class from the /usr/lib/libstdc++.so.5 library instead and as a result DRLVM fails. I'm going to continue this investigation and prepare a patch to fix this issue if someone doesn't already know how. Volunteers? Thanks, Vladimir. On 7/18/06, Andrey Chernyshev <[EMAIL PROTECTED]> wrote:
On 7/17/06, Vladimir Gorr <[EMAIL PROTECTED]> wrote: > On 7/17/06, Salikh Zakirov <[EMAIL PROTECTED]> wrote: > > > > Hi, > > > > Can anyone share their experience on running DRLVM built from SVN? > > > > On my workstations, it requires several fixes in order to run properly: > > > > HARMONY-898 'workaround to get correct hythr.dll' (both for Linux and > > Windows) > > HARMONY-853 'DRLVM+classlib segfaults in hyzlib' (on Linux) > > > > Without HARMONY-898, DRLVM segfaults after reading NULL from TLS. > > Without HARMONY-853, DRLVM segfaults in libhyzlib.so on Linux. > > > > I would suggest that the patches from HARMONY-898 and HARMONY-853 be > > applied, > > because DRLVM doesn't work at all for me without these fixes. > > > > Any other experiences? Is anyone besides me running DRLVM from SVN? > > > > I've built the DRLVM on Linux for the following cases: > > 1. w/o patches listed above; > > 2. w/ H-898 patch; > > 3. w/ H-898 & H-853 patches. > > > > 'Segmentation fault' occurs for all cases, unfortunately. > I was able to run most of smoke tests on Linux a few days ago, but now I'm having the same segfault failure. Some observations are: - In the interpreted mode (-Xint option) it reports ClassNotFoundException, so it seems VM can no longer load the classes from jars on Linux. I was able to run a single "Hello" class though. - Segfault actually happens during VM shutdown, after the main() is already completed with CNFE error. I wasn't able to find the exact point of segfault though - it seems it happens somewhere in a jitted code, gdb doesn't show a much of useful info in that case: #0 0x40788aa2 in malloc_consolidate () from /lib/tls/libc.so.6 #1 0x00000010 in ?? () #2 0x4083c7c8 in main_arena () from /lib/tls/libc.so.6 #3 0x00000000 in ?? () #4 0x4083c7b4 in main_arena () from /lib/tls/libc.so.6 #5 0x4083c798 in main_arena () from /lib/tls/libc.so.6 #6 0x4083c780 in __malloc_initialize_hook () from /lib/tls/libc.so.6 #7 0x4083aff4 in ?? () from /lib/tls/libc.so.6 #8 0x0855cdd0 in ?? () #9 0x00037230 in ?? () #10 0x56500158 in ?? () #11 0x40789bf6 in _int_free () from /lib/tls/libc.so.6 Previous frame inner to this frame (corrupt stack?) I'm continuing investigation, the same behavior is observed on Linux if run with classlib's launcher. Thanks, Andrey. > > > Thanks, > > Vladimir. > > --------------------------------------------------------------------- > > Terms of use : http://incubator.apache.org/harmony/mailing.html > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > -- Andrey Chernyshev Intel Middleware Products Division --------------------------------------------------------------------- Terms of use : http://incubator.apache.org/harmony/mailing.html To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
