regex_regress fails under intel-linux-10.0
Hi, This test, regex_regress, is failing under intel-10 with this error message. [kbelco@sahp7645 results]$ head /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output terminate called without an active exception terminate called recursively terminate called recursively terminate called recursively terminate called recursively terminate called recursively terminate called recursively terminate called recursively terminate called recursively terminate called recursively [ ... ] Where the "terminate called recursively" is repeated approximately 570e6 times. [kbelco@sahp7645 results]$ ls -l /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output -rw-rw-r-- 1 kbelco kbelco 572424926 Dec 27 04:58 /var/scratch2/ boost/results/boost/bin.v2/libs/regex/test/regex_regress.test/intel- linux-10.0/debug/threading-multi/regex_regress.output This is causing process_jam_log to fail to process the intel-10.0 results, resulting in a hang. Any idea why this test is failing under intel-10.0 while it appears to pass okay under 8.1, 9.0 and 9.1? Thanks. -- Noel Belcourt [ Here's the bjam.log output for this test ] testing.capture-output /var/scratch2/boost/results/boost/bin.v2/libs/ regex/test/regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.run LD_LIBRARY_PATH=/usr/local/intel/cc/10.0.025/lib:/var/scratch2/ boost/results/boost/bin.v2/libs/regex/build/intel-linux-10.0/debug/ threading-multi:$LD_LIBRARY_PATH export LD_LIBRARY_PATH "/var/scratch2/boost/results/boost/bin.v2/libs/regex/test/ regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress" > "/var/scratch2/boost/results/boost/bin.v2/libs/ regex/test/regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.output" 2>&1 status=$? echo >> "/var/scratch2/boost/results/boost/bin.v2/libs/regex/ test/regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.output" echo EXIT STATUS: $status >> "/var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output" if test $status -eq 0 ; then cp "/var/scratch2/boost/results/boost/bin.v2/libs/regex/test/ regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.output" "/var/scratch2/boost/results/boost/bin.v2/libs/ regex/test/regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.run" fi verbose=0 if test $status -ne 0 ; then verbose=1 fi if test $verbose -eq 1 ; then echo ====== BEGIN OUTPUT ====== cat "/var/scratch2/boost/results/boost/bin.v2/libs/regex/ test/regex_regress.test/intel-linux-10.0/debug/threading-multi/ regex_regress.output" echo ====== END OUTPUT ====== fi exit $status 300 second time limit exceeded ...failed testing.capture-output /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.run...
K. Noel Belcourt wrote:
Hi,
This test, regex_regress, is failing under intel-10 with this error message.
Is this Linux or Win32? Passes OK for me here with Intel-10 on Win32.
[kbelco@sahp7645 results]$ head /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output
terminate called without an active exception
That looks very odd to me? How can that ever happen? Can you give me the platform details so I can try and reproduce it, if you can try running under gdb and see if you can get a backtrace that might help too. Cheers, John.
K. Noel Belcourt wrote:
Where the "terminate called recursively" is repeated approximately 570e6 times.
[kbelco@sahp7645 results]$ ls -l /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output -rw-rw-r-- 1 kbelco kbelco 572424926 Dec 27 04:58 /var/scratch2/ boost/results/boost/bin.v2/libs/regex/test/regex_regress.test/intel- linux-10.0/debug/threading-multi/regex_regress.output
This is causing process_jam_log to fail to process the intel-10.0 results, resulting in a hang. Any idea why this test is failing under intel-10.0 while it appears to pass okay under 8.1, 9.0 and 9.1?
I tried this on my Linux box with both Intel-10.0 and Intel-10.1 and see similar behaviour in each case: the program hangs (actually it segfaults according to gdb) as soon as it attempts to throw an exception from within the library. However, I'm using this compiler on an unsupported Linux configuration, so problems like this aren't necessarily unexpected :-( So... is the system you're running on supported officially by Intel? Also like you I have no issues with earlier Intel compiler releases on this machine, so I'm not sure what if anything I can do to fix this :-( Just an idea: but can you try building that test with link=static and see if that helps at all? Just in case it's a passing exceptions across .so boundaries problem? Thanks, John.
Hi John, On Dec 28, 2007, at 4:05 AM, John Maddock wrote:
K. Noel Belcourt wrote:
Where the "terminate called recursively" is repeated approximately 570e6 times.
[kbelco@sahp7645 results]$ ls -l /var/scratch2/boost/results/boost/ bin.v2/libs/regex/test/regex_regress.test/intel-linux-10.0/debug/ threading-multi/regex_regress.output -rw-rw-r-- 1 kbelco kbelco 572424926 Dec 27 04:58 /var/scratch2/ boost/results/boost/bin.v2/libs/regex/test/regex_regress.test/intel- linux-10.0/debug/threading-multi/regex_regress.output
This is causing process_jam_log to fail to process the intel-10.0 results, resulting in a hang. Any idea why this test is failing under intel-10.0 while it appears to pass okay under 8.1, 9.0 and 9.1?
I tried this on my Linux box with both Intel-10.0 and Intel-10.1 and see similar behaviour in each case: the program hangs (actually it segfaults according to gdb) as soon as it attempts to throw an exception from within the library.
However, I'm using this compiler on an unsupported Linux configuration, so problems like this aren't necessarily unexpected :-(
So... is the system you're running on supported officially by Intel?
From what I can tell, yes. The only issue I see is that Intel's web site recommends using libstdc++.so.5 and we have both libstdc++.so versions 5.0.7 and 6.0.3. I suppose it could be an issue with using version 6.0.3.
Also like you I have no issues with earlier Intel compiler releases on this machine, so I'm not sure what if anything I can do to fix this :-(
Just an idea: but can you try building that test with link=static and see if that helps at all? Just in case it's a passing exceptions across .so boundaries problem?
With link=static, I reran the entire set of intel-linux-10 tests. The tests ran to completion and the regex_regress test passed. I'll start posting intel-linux-10.0 results for the trunk with link=static. Thanks for the help. -- Noel
participants (2)
-
John Maddock -
K. Noel Belcourt