| Ticket UUID: | 1914388 | |||
| Title: | test case failures under windows xp (sp2) | |||
| Type: | Bug | Version: | None | |
| Submitter: | melbardis | Created on: | 2008-03-14 15:50:34 | |
| Subsystem: | 43. Regexp | Assigned To: | msofer | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2008-07-21 19:14:10 | |
| Resolution: | None | Closed By: | patthoyts | |
| Closed on: | 2008-07-21 12:14:10 | |||
| Description: |
I tried the following builds on Windows XP (sp2) and ran the test cases with the following results. CC is msvc 13.10.6030 .net 2003 cygwin build tools (attached is a zip file with full results) 1) ./configure CC=cl Files with failing tests: chanio.test http.test httpold.test Test files exiting with errors: reg.test 2) ./configure CC=cl --enable-threads Files with failing tests: http.test httpold.test tcltest.test Test files exiting with errors: reg.test 3) ./configure CC=cl --enable-symbols Files with failing tests: http.test httpold.test Test files exiting with errors: reg.test 4) ./configure CC=cl --enable-symbols --enable-threads Files with failing tests: http.test httpold.test Test files exiting with errors: reg.test Re-ran #1 to see if the chanio.test failure was fluke, it was not... (more tests failed!) Tcl2.log Files with failing tests: chanio.test http.test httpold.test io.test tcltest.test Test files exiting with errors: reg.test Note: I did a make clean between the tests | |||
| User Comments: |
patthoyts added on 2008-07-21 19:14:10:
Logged In: YES user_id=202636 Originator: NO there are now no differences in the failing tests between msys and nmake built versions. msofer added on 2008-07-21 11:45:56: Logged In: YES user_id=148712 Originator: NO Still happening on HEAD? msofer added on 2008-06-28 20:38:51: Logged In: YES user_id=148712 Originator: NO reassigning to keep following 1905562 kennykb added on 2008-03-24 21:05:26: Logged In: YES user_id=99768 Originator: NO reassigning this one to follow 1905562. It happens that pat is right, because the Makefile.vc and the autoconf-generated Makefile happen to configure the executable with different stack sizes. Resizing the regexp accordingly brings back the crash. If we want to cure this one for real, we probably either need to go to stackless within the regex engine (dream on?) or else put the stack check on the potentially unbounded recursion there. patthoyts added on 2008-03-24 19:18:25: Logged In: YES user_id=202636 Originator: NO These are side effects of the build environment. If Tcl HEAD is built using the MSVC .NET 2003 compiler but the tests are run just under the cmd.exe shell then they all pass. eg: nmake -f makefile.vc OPTS=thread test dgp added on 2008-03-14 23:14:53: Logged In: YES user_id=80530 Originator: NO use the -verbose option to determine which test(s) are causing segfaults. assigning to kbk in the likelihood this is a Dup of 1905562 melbardis added on 2008-03-14 22:50:35: File Added - 270411: log.zip | |||
Attachments:
- log.zip [download] added by melbardis on 2008-03-14 22:50:35. [details]