Ticket UUID: | 2996549 | |||
Title: | Failure in expr.test on Win32 | |||
Type: | Bug | Version: | obsolete: 8.6b1.1 | |
Submitter: | nijtmans | Created on: | 2010-05-04 12:46:54 | |
Subsystem: | 48. Number Handling | Assigned To: | kennykb | |
Priority: | 5 Medium | Severity: | ||
Status: | Closed | Last Modified: | 2011-04-08 13:49:20 | |
Resolution: | Fixed | Closed By: | nijtmans | |
Closed on: | 2011-04-08 06:49:20 | |||
Description: |
Running the test suite with HEAD, gives the following result. This tests did not fail before, so it must be recently introduced. =========================================== Tests began at Tue May 04 14:27:33 +0200 2010 expr.test ==== expr-29.1 smallest representible number FAILED ==== Contents of test case: list [catch {convertToDouble 4.9406564584124654e-324} result] $result [catch {convertToDouble 2.4703282292062327e-324} result] $result [catch {convertToDouble 2.47032822920623e-324} result] $result ---- Result was: 0 0x0000000000000001 0 0x0000000000000000 0 0x0000000000000000 ---- Result should have been (exact matching): 0 0x0000000000000001 0 0x0000000000000001 0 0x0000000000000000 ==== expr-29.1 FAILED ==== expr-29.2 smallest representible number FAILED ==== Contents of test case: list [catch {convertToDouble -4.9406564584124654e-324} result] $result [catch {convertToDouble -2.4703282292062327e-324} result] $result [catch {convertToDouble -2.47032822920623e-324} result] $result ---- Result was: 0 0x8000000000000001 0 0x8000000000000000 0 0x8000000000000000 ---- Result should have been (exact matching): 0 0x8000000000000001 0 0x8000000000000001 0 0x8000000000000000 ==== expr-29.2 FAILED Tests ended at Tue May 04 14:27:37 +0200 2010 all.tcl:Total1865Passed1856Skipped7Failed2 | |||
User Comments: |
nijtmans added on 2011-04-08 13:49:20:
allow_comments - 1 nijtmans added on 2010-05-19 18:38:48: No need to backport to 8.5, because the bug is not there: The fix for [Bug 2952904] introduced this new bug, but this fix was only applied in HEAD. Once #2952904 is backported, #2996549 should be backported along with it. dgp added on 2010-05-18 20:50:29: thanks. any reason not to backport this for 8.5.9 ? nijtmans added on 2010-05-18 04:52:03: Fixed in HEAD nijtmans added on 2010-05-04 20:38:54: File Added - 372801: 2996549-2.patch nijtmans added on 2010-05-04 20:38:28: Here is a better fix, which delays the initialization of "tiny". So it fixes this bug without re-introducing [Bug 2952904] (I hope). Please evaluate. nijtmans added on 2010-05-04 20:24:18: Reverting the change from tclStrToD.c from 1.41 to 1.42 (attached patch) fixes it. Of course, that would re-introduce [Bug 2952904], so something smarter is needed. nijtmans added on 2010-05-04 20:21:58: File Added - 372797: 2996549.patch nijtmans added on 2010-05-04 20:01:30: A binary search points to the following change which started this test case to fail, therefore assigning to Kevin 2010-04-02 Kevin B. Kenny <[email protected]> * generic/tclStrToD.c: [Bug 2952904]: Defer creation of the smallest floating point number until it is actually used. (This change avoids a bogus syslog message regarding a 'floating point software assist fault' on SGI systems.) dgp added on 2010-05-04 19:56:16: Can you bracket working vs. non-working? Start with reporting whether the 8.6b1 release works in the same environment. See also 1618447. |