Opened 14 years ago
Closed 14 years ago
#3279 closed defect (fixed)
PCIDSK2 fails on misc_5 on Windows slavebots (only x64 ones?)
| Reported by: | Even Rouault | Owned by: | warmerdam |
|---|---|---|---|
| Priority: | normal | Milestone: | 1.7.0 |
| Component: | Buildbot | Version: | unspecified |
| Severity: | normal | Keywords: | pcidsk |
| Cc: |
Description
Frank,
I think I've tracked down the failures on Windows slavebots to the case where we try to create a PCIDSK2 dataset with nBands == 0
First try = All tests enabled --> fails : http://buildbot.osgeo.org:8500/builders/szekerest-vc80x64-full/builds/239/steps/test/logs/stdio (the print traces are probably truncated a bit before the actual failure)
Second try = All test enabled, PCIDSK misc_5 tests disabled --> succeeds : http://buildbot.osgeo.org:8500/builders/szekerest-vc80x64-full/builds/240/steps/test/logs/stdio
Third try = All test enabled, PCIDSK misc_5 tests enabled, excepted the case where nBands == 0 --> succeeds: http://buildbot.osgeo.org:8500/builders/szekerest-vc80x64-full/builds/241/steps/test/logs/stdio
A possible suspect : &(aeChanTypes[0]) in pcidskdataset2.cpp ? (it might be actually correct. I just don't know how std::vector behaves when trying to do that on empty vector)
Change History (1)
comment:1 by , 14 years ago
| Component: | default → Buildbot |
|---|---|
| Keywords: | pcidsk added |
| Resolution: | → fixed |
| Status: | new → closed |
I think this is fixed in trunk (r18295) and I have re-enabled the misc.py test (r18297).
It seems reasonable that taking a reference to a non-existant array might be undefined.