Opened 8 years ago
Last modified 7 years ago
#64 accepted defect
VM crashes in fclose()
| Reported by: | vranyj1 | Owned by: | vranyj1 |
|---|---|---|---|
| Priority: | major | Milestone: | |
| Component: | virtual machine | Keywords: | |
| Cc: | Also affects CVS HEAD (eXept version): | yes |
Description
The Linux 32bit VM crashes (SEGV) in fclose() when running attached script.
It may or may not be another incarnation of "double-fclose" of "fseek-after-fclose" bugs.
Attachments (1)
Change History (4)
Changed 8 years ago by
| Attachment: | bug_55_crash_on_fclose_2.st added |
|---|
comment:1 Changed 7 years ago by
| Owner: | set to vranyj1 |
|---|---|
| Status: | new → accepted |
comment:2 Changed 7 years ago by
Hmm...this testcase is still broken on Windows 32bit. It manifests in a funny way - it truncates files it opens, i.e., makes them empty. After some interations, some of the opened .st files are empty.
Based on procmon from SysInternals? Suite, a CloseHandle() is called twice on some handles - and those files are truncated. This makes me thinking there's yet some unfixed
problem in weak arrays.
comment:3 Changed 7 years ago by
This is tested by test #test_issue_64_vm_crashes_in_fclose. Commit dbd517ea6e1c/stx.goodies.regression disables this test on Windows.
As of today, the bug seems to be automagically fixed. Attached testcase does not crash
the VM after some 300 iterations on 32bit and 800 iterations on 64bits. It may be that some C99-fixes in the librun plus -fwrapv options somehow "fixed" it.