Bugs item #1895125, was opened at 2008-02-16 23:13
Message generated for change (Comment added) made by tsheyar
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1895125&group_id=56967

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: PF/compiler
Group: Pathfinder 0.22
Status: Open
Resolution: None
Priority: 6
Private: No
Submitted By: Stefan Manegold (stmane)
Assigned to: Manuel Mayr (mayrm)
Summary: PF: 28 ALG tests still segfault

Initial Comment:
After Manuel's fix of

[ 1894964 ] PF/ALG: 170 tests segfault in atom_eq() in opt_algebra_cse.c
http://sourceforge.net/tracker/index.php?func=detail&aid=1894964&group_id=56967&atid=482468

28 of the 170 tests still segfault in various places and/or for various reasons.

See `Mtest.py -G ...` and/or the attached list and log for details.


----------------------------------------------------------------------

>Comment By: Jan Rittinger (tsheyar)
Date: 2008-02-19 17:26

Message:
Logged In: YES 
user_id=993208
Originator: NO

It turned out that line 585 in algebra/algebra_cse.c <code>a = *(PFarray_t
**) PFarray_at (subexps, n->kind);</code> relies on the fact that memory is
allocated with calloc instead of malloc. This triggered queries
ALGsegfault.SF-1894964{a,b} to fail on some platforms. (Similar problems
arise in algebra/opt/opt_algebra_cse.c.)

Note that this is a general problem of all scenarios where PFarray is used
as an array instead of a list.


----------------------------------------------------------------------

Comment By: Stefan Manegold (stmane)
Date: 2008-02-19 13:20

Message:
Logged In: YES 
user_id=572415
Originator: YES

2.)
While working fine with MPS on all platforms
(cf.,
http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/GNU.64.64.d-Fedora8/tests_BugTracker/ALGsegfault.SF-1894964a.err.00.html
http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/GNU.64.64.d-Fedora8/tests_BugTracker/ALGsegfault.SF-1894964b.err.00.html
),
the segfault with the third query occurs with ALG only on 64-bit x86_64
machines running Fedora (6 & 8)
(cf.,
http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTestsG103/GNU.64.64.d-Fedora8/tests_BugTracker/ALGsegfault.SF-1894964a.err.00.html
http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTestsG103/GNU.64.64.d-Fedora8/tests_BugTracker/ALGsegfault.SF-1894964b.err.00.html
);
it works fine also with ALG on 32-bit systems as well as on 64-bit Windows
(on x86_64) and on 64-bit RedHat (on IA-64 (Itanium)) ...


----------------------------------------------------------------------

Comment By: Stefan Manegold (stmane)
Date: 2008-02-18 19:56

Message:
Logged In: YES 
user_id=572415
Originator: YES

Jan,

thanks for the fixes and analysis!

1.)
We should conditionally disable (in the XQuery_0-22 release branch, only)
all tests that are known/bound to fail due to not yet (fully) available
features; see also
[ 1895322 ] PF/alg: 20 tests fail with (various) "fatal error"
http://sourceforge.net/tracker/index.php?func=detail&aid=1895322&group_id=56967&atid=482468

2.)
tests/BugTracker/Tests/ALGsegfault.SF-1894964{a,b}.*; still fails for me
on my 64-bit Fedora8 desktop (both with "--enable-debug --enable-assert"
and "--enable-optimize --disable-assert) --- let's see what tonights
testing will reveal tomorrow morning; might be a 64-bit and/or
platform-specific problem ...

3.)
See 1.) above.


----------------------------------------------------------------------

Comment By: Jan Rittinger (tsheyar)
Date: 2008-02-18 14:51

Message:
Logged In: YES 
user_id=993208
Originator: NO

1.) By producing an error message for the missing self step (instead of an
assertion) we now have less segfault queries (and of course more with a
FATAL ERROR.

2.) tests/BugTracker/Tests/ALGsegfault.SF-1894964{a,b}.*; works as
expected on my setup

3.) Various queries listed in ALG_segfault_2 do not segfault on my machine
anymore, but produce only an error message (mostly because of missing
features). In fact all queries for which you attached a gdb profile now do
not segfault anymore.

Stefan, what do you think we should do with this bug report? -- Would you
call it fixed?

----------------------------------------------------------------------

Comment By: Stefan Manegold (stmane)
Date: 2008-02-18 07:35

Message:
Logged In: YES 
user_id=572415
Originator: YES

Some of the segfaults with ALG occur because running more than 2 (two)
queries
using the ALG version on the same `Mserver
--dbinit="module(pathfinder);"`
make the Mserver segfault in "subexp_eq()" in
"compiler/algebra/algebra_cse.c" .

E.g., see newly added test
tests/BugTracker/Tests/ALGsegfault.SF-1894964{a,b}.*;
simply compare
`Mtest.py -q tests/BugTracker/ ALGsegfault.SF-1894964{a,b}`
and
`Mtest.py -G -q tests/BugTracker/ ALGsegfault.SF-1894964{a,b}`

See the attached log for details.

File Added: ALG_segfault_3_GDB_log

----------------------------------------------------------------------

Comment By: Stefan Manegold (stmane)
Date: 2008-02-16 23:16

Message:
Logged In: YES 
user_id=572415
Originator: YES

File Added: ALG_segfault_2_GDB_log

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1895125&group_id=56967

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Monetdb-bugs mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/monetdb-bugs

Reply via email to