FileFormatTest: Provide data to MOC for big endian tests#2479
FileFormatTest: Provide data to MOC for big endian tests#2479pkturner wants to merge 1 commit intoOpenOrienteering:masterfrom
Conversation
The ocdTextSymbolTest and ocdAreaSymbolTests had been created with 0 rows on big endian systems, which was causing a fatal error in QTest::fetchData.
|
Thank you @pkturner. |
That's because NetCDF is not built for Z, therefore GDAL does not build. As a result, anything depending on GDAL does not build on Z either. I'm looking into a fix. I've looked if other distros, .deb or .rpm based, can build Mapper on s390x, but all seem to be broken. Tumbleweed appears to be the easiest to fix. As a side note, it's reportedly possible to gain access to System Z through LinuxONE project. I haven't tried LinuxONE myself yet, but if someone is interested in first-hand experience with that platform, it might be a way. Spoiler alert -- over SSH it looks like any other Linux instance. |
|
GDAL builds now for s390x in my staging. I've created a custom build of pkturner's submission in https://build.opensuse.org/package/show/home:LPechacek:oomapper/openorienteering-mapper-master (updated _service file to point to his Git repo). The test results look better now but there are still glitches: |
|
Thank you @lpechacek for testing my PR. Those glitches were present in issue #2473 and escaped my notice. I will work on a change to make them EXPECTed FAILs. |
This addresses #2473 (file_format_t fails on s390x). The
ocdTextSymbolTestandocdAreaSymbolTestshad been created with 0 rows on big endian systems, which was causing a fatal error inQTest::fetchData.The
understandsTesttest already had a technique for skipping OCD test cases on big endian systems, so I applied the same in theocdTextSymbolTestandocdAreaSymbolTests, after removing the#ifdefs which were causing failure.I don't have a big endian system at my disposal, so I forced the definition of the
MAPPER_BIG_ENDIANmacro, first to duplicate the fatal error, then to validate the code change.