|
|
Boost Testing : |
From: Robert Ramey (ramey_at_[hidden])
Date: 2008-06-19 14:08:34
a) Comeau compiler complains about NULL not being defined. I used the
the previous version of comeau without problems in this area. I've NULL
thoughout the code and never had a problem with it not being defined.
I realise that views on NULL and what it means - or should mean -
if anything have evolved over the years - and I don't want to go
into that here. I doubt the serialization library is unique in this
aspect. Is there anything that can be done in the testing setup/
configuration to address this?
b) RudbekAssociates-V2 and Huang-Vista-x64 are still showing
results from tests which were removed from the test suite months
ago. Worse, the web format make them look like current
results. Is there not anything that can be done to fix this?
c) msvc-9.0-x64-stdcxx-4.2.1 and msvc-7.1-stdcxx-4.2.1 show
test failure but when I follow the link - I get "file not found".
This is not helpful to me.
I made changes to fix previous failures on this combination
and I think they are correct - but these tests aren't
confirming or disproving this.
d) borland- 5.9.3 shows failure to compile the library.
But when I follow the link - I get some text output
which shows only a warning for one code module.
e) borland 5.6.4 shows failure to build the library.
When I follow the link, I get compile succed on
one file. I doubt this is correct. I have no idea
what to do about this.
f) Intel compilers on Darwin compiling this
wide character versions emits warnings
on spirit compile. Then a cryptic message
regarding linking symbols. Any one able
to shed light on this?
g) HP compilers with wide character compiles
fail due to failure of the test for wide character
support in config tests. It would be surprising
to me that this platform wouldn't supply wstream
i/o. Could someone interested in this platform
investigat why config/wstreab_buf test is failing?
h) Recent changes in "export" functionality have
resulted in much cleaner and effective code. Unfortunately,
in the shuffle, it seems something broke in this
area for the Intel compiler. (Maybe the SUN
compiler as well). This shows up in a failure of
test_exported. Files involved are export.hpp and
perhaps force_include.hpp. If someone with interest
in the intel compiler and serialization were to spend
some time investigating and experimenting with this, it would be much
appreciated. I don't have this compiler and intel
doesn't seem to want to give me a free one, so
I can't do this myself - even assuming I had time
to spend on this.
Aside from the above, I have a few failures left
that have left me puzzled, but I don't see anyone
can help me with them.
Robert Ramey