This is not a bug in the Test Suite but rather a workaround for CMake/Valgrind under MacOS X (I should report this to either or both of those projects as well).
Using CMake 2.8.12.2 & valgrind 3.10.0, md5 checksum hangs. It would even be better if the md5 check wasn't run under valgrind but I don't know if it can be done (search for a while but didn't find any test property useful for this).
Attached is a patch that works around this problem. It prevents another cmake process being run & use file(MD5 instead. Tested in CDash.
- md5.patch 1.07KB
Comment #1
Posted on Oct 22, 2014 by Happy Rabbit- cc Mathieu Malaterre
@Mathieu: could you have a look to this issue ?
Comment #2
Posted on Oct 22, 2014 by Helpful GiraffePlease do not apply the suggested patch, as explained by OP this is a hack. It solves the symptoms not the bug, plus it removes essential features of the test suite.
I do not see why running cmake -E md5sum
would make valgrind hang...but I do not have access to a MacOSX box ATM
Comment #3
Posted on Oct 22, 2014 by Quick Lion@Mathieu,
The way MD5 is computed is changed, that's all. It doesn't remove any essential features from the test suite as you can see on CDash (mbpro-mlr has this patch).
IMHO, while achieving non regression, it makes the file checkmd5refs.cmake easier to read. One could argue that the original is a hack, I asked myself why a process was spawned there... maybe if "file(MD5" was buggy & not cmake -E md5sum
in some version of CMake.
Comment #4
Posted on Oct 22, 2014 by Helpful Giraffe@matthieu ok my mistake I missed the file(MD5 ${pgxfullpath} output)
line.
It makes sense to remove this from the test suite, indeed. Patch can applied as-is. thx for the clarification.
Comment #5
Posted on Oct 22, 2014 by Happy RabbitThis issue was closed by revision r2914.
Status: Fixed
Labels:
Type-Defect
Priority-Medium