Ok, here are how I tested and my results:
1. On my Mac (OSX 10.3.9) I cleared out chandler/release/* to make sure
I don't have any leftover crud from yesterday's tests, then in chandler
did "svn up; make install" to get the current everything for the
application and also all the release/* stuff.
2. In external/swig made sure that I have the latest and not the 1.3.24
version I was playing with yesterday, then did "make DEBUG=0"
3. Did a complete new checkout of internal/wx, again to ensure that
there were not any leavings from prior tests, using svn rev 9544 (which
was when David bumped the relver to 36.)
4. In internal/wx I did "make DEBUG=0". Because of my trick of
sym-linking the external/release dir to chandler/release this will
install the wx libs and wxPython modules to chandler/release saving me
some steps.
5. Back over to ../../chandler and made sure chandler would start
normally. Then ran the functional tests three times. Each run passed
all tests with no errors, although I did get all of the "swig detected a
memory leak" errors, but these are just noise since I only used the
trick that causes this message in the new swig on objects that get owned
by their parent. The changes needed to suppress this message (i.e. to
handle ownership transfer the new "correct" way in swig) are part of the
things that have been done since -36.
6. Did the same on Linux with the same results, although I had to make
sure that Chandler remained the foreground app, otherwise some of the
test scripts would fail. I assume that they are doing something like a
wx.Window.FindFocus and failing when they try to use the None return
value as if it was a widget.
Summary:
Yesterday (current wx, swig-1.3.24):
FT's failed on OSX, passed on Linux
Today (wx-36, swig-1.3.29):
FT's passed on OSX and Linux
So it looks like SWIG isn't the culprit, unless I just got lucky and the
totally random failures didn't fail for me three times in a row.
Robin
John Anderson wrote:
The best way I've been able to reproduce the problem is to run the
functional tests on Mac release. However, the failures are random and
the tests sometimes succeed.
John
Robin Dunn wrote:
Philippe Bossut wrote:
There's also the SWIG issue. May be Robin should try to build a
version of Chandler with wx 36 and the new SWIG so that we can see if
this new version of SWIG is indeed the cause of some of the problems.
It's building... Is there anything besides the Functional Tests that
I should do to try and reproduce the problems? Or is it totally random?
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
