harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ilya Berezhniuk (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HARMONY-5262) [drlvm][gc_cc] GC_CC does not work in debug mode since r583223
Date Sat, 08 Dec 2007 00:46:43 GMT

     [ https://issues.apache.org/jira/browse/HARMONY-5262?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Ilya Berezhniuk updated HARMONY-5262:
-------------------------------------

    Attachment: gc_cc_0005.patch

Here is the updated patch.
The patch was tested together with HARMONY-5247 patch to get rid of hangups.

Command looked like
./build.sh -Dtest.vmargs="-XX:-vm.assert_dialog -XX:gc.dll=libgc_cc.so" test

With this patch I've passed 'build test' with GC_CC on Linux/Windows x86, and on Linux/x86_64.

Windows/x86_64 showed the following intermittent failures during several runs (other tests
are passed):
outofmemory.Int (in JIT mode)
outofmemory.Plain (in JIT mode)
perf.StressCornerSize (in JIT mode)
stress.Mix (in OPT mode)

I've reproduced these failures in cyclic runs of particular tests. There are strange assertions
like
----------------
hythread_suspend_disable()
.....
assert(!hythread_is_suspend_enabled())
---------------
The code between these lines does not change disable_count (problems with cache/atomicity?).


I'm going to try some additional changes, but I'm not sure in success. Anyway this patch looks
useful.
This fix affects only GC_CC; 'build test' with GC_GEN is passed on all 4 tested platforms.


> [drlvm][gc_cc] GC_CC does not work in debug mode since r583223
> --------------------------------------------------------------
>
>                 Key: HARMONY-5262
>                 URL: https://issues.apache.org/jira/browse/HARMONY-5262
>             Project: Harmony
>          Issue Type: Bug
>          Components: DRLVM
>         Environment: All
>            Reporter: Ilya Berezhniuk
>         Attachments: gc_cc_0003.patch, gc_cc_0005.patch
>
>
> I've tried to pass 'build test' with GC_CC and got a lot of failures, and also several
random hangs.
> When investigating these problems, I've detected that hang occurs in shutdown (HARMONY-5247).
> But tests hanged with GC_CC on M3 snapshot (I've checked this).
> I've found that failures are caused by lots of assertions introduced by threading changes
from HARMONY-4830.
> The attached patch fixes GC_CC to reflect threading changes.
> It also corrects GC_CC according to r599482 which included changes in object layout.
> With this patch the tests still often hang (because of HARMONY-5247).
> But almost all 'build test' tests pass for me on Windows when they do not hang.
> I still have a problem with loading gc_cc library on Linux (tl::MemoryPoolMT can't be
found).
> Also outofemory.Int still fails on x86_64.
> I'll try to fix these remaining problems and update the patch.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message