Fixed -O -inline and -release incompatibilty with Orange. #36
Conversation
… for Windows. Exactly the same as unittest.sh, but ported to a batch file.
… unittests_and_release
There was a problem hiding this comment.
No, but tests failed when -O -inline was added to unittest.sh (and there was an obscure comment in the Makefile's). The output was the following:
orange.serialization.SerializationException.SerializationException@orange\serial
ization\SerializationException.d(25): Could not find an element "array" with the
attribute "key" with the value "arr".
----------------
0x004273EA in D6orange13serialization10Serializer10Serializer6__ctorMFC6orange13
serialization8archives7Archive7ArchiveZC6orange13serialization10Serializer10Seri
alizer9__lambda2MFNaNfC6orange13serialization22SerializationException22Serializa
tionExceptionZv
----------------
when compiling with -g -O -inline -release.
--jm
Fixed Primitive.d and added to pull request. |
|
Could you please squash these commits in to one. Or perhaps two, one for the Windows unit test file and one for the rest. |
|
No problem, I'll create one branch for the unittest.bat file and |
|
I'm a little busy today, I won't have time to separate the commits until tonight. |
|
No rush. |
…unittests_and_release
|
Are the commits squashed? I would prefer to not have an extra merge commit. Could you please do a rebase instead of a merge? |
2 participants
Fixed -O -inline and -release incompatibilty with Orange. (inlining was omitting the 'else' block of the static ifs in toData and fromData).
I really needed to make it work with -O -inline, and I'm happy that I've found the problems. If there is a better way, please use that instead.
Also added an automatic "unittest.bat" runner for Windows ("ported" unittest.sh, gathering *.d in the same way in a batch file).
The "Primitive" test now passes in Windows, though the FP string zero representation should actually be portable? (Maybe I have a locale difference? "0x0.p+0" vs "0x0p+0").
--Juan Manuel Cabo