Re: Testing issues

Vladimir Prus wrote:
> Or we can approach the problem from a different angle
and implement
> some special hook that stops "threading" from
showing up in
> paths, just for testing.

As we talked about in IRC, I went ahead and implemented
removing of the 
features subdir parts in BoostBuild/py. This makes more
tests pass on 
MSVC/Windows. Unfortunately we now have expect_content tests
which don't 
pass because it fails to open, as the path is wrong, the
generated files:

===txt
project_test3             :  Note: could not open file 
bin/msvc-8.0*/debug/a.obj
-------- all changes caused by last build command
----------
Added files   :
['lib/bin/msvc-8.0/debug/threading-multi/b.obj', 
'lib/bin/msvc-8.0/debug/threading-multi/m.exe', 
'lib2/helper/bin/msvc-8.0/debug/threading-multi/e.obj', 
'lib2/bin/msvc-8.0/debug/threading-multi/c.obj', 
'lib2/bin/msvc-8.0/debug/threading-multi/d.obj', 
'lib2/bin/msvc-8.0/debug/threading-multi/l.exe', 
'lib3/bin/msvc-8.0/debug/threading-multi/f.obj', 
'bin/msvc-8.0/debug/threading-multi/a.exe', 
'bin/msvc-8.0/debug/threading-multi/a.obj']
Removed files : []
Modified files: []
Touched files : []
STDOUT ============

STDERR ============

END ===============
FAILED
===

This going to take a bit more trickery. We'll have to build
a map so we 
can remove the threading-multi subdir, match against the
file being 
opened, and map back to the real generated file.


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software
.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build

Rene Rivera wrote:
> This going to take a bit more trickery. We'll have to
build a map so we 
> can remove the threading-multi subdir, match against
the file being 
> opened, and map back to the real generated file.

Or something like that... A bit more changes and fixes in
other context 
under which a directory structure is assumed and most tests
pass on 
MSVC/Windows. Tests that fail:

rebuilds [1]
generators_test [2]
searched_lib
alternatives
inline [2]
tag [2]
project_glob
regression [3]
project_root_rule [2]
remove_requirement [1,4]
gcc_runtime [1,5]
pch

1. Also fails on GCC/Linux
2. This is the difference of Windows generating lib*.lib and
the test 
expecting *.lib
3. EOL discrepancy of some sort on the expected vs. actual
output
4. Removing the "threading-multi" hurts us on this
test.
5. Fails on both for rather different reasons.

Another testing difference between Linux and Windows: the
output of 
exceptions is not inlined the same way. On Windows it seems
to be 
buffered, and hence looks like it comes out garbled. But
it's really 
just in rather different places than when the exceptions are
actually 
thrown/printed.


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software
.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build