|
Boost : |
From: David Abrahams (dave_at_[hidden])
Date: 2005-05-26 14:42:52
"David Daeschler" <daveregs_at_[hidden]> writes:
> Thanks to that output (-a -odump.txt ), I found something interesting when
> doing testing.
>
> When you run the tests on the boost filesystem, it builds a static library
> at:
>
> bin\boost\libs\filesystem\build\libboost_filesystem.lib\mingw\debug\libboost_filesystem-mgw-d-1_32.lib
>
> It then deletes the library when the tests are over.
That's so that when we're doing the big regression tests we don't fill
up testers' disks needlessly. You can pass --preserve-test-targets to
bjam to suppress that behavior.
> If I break the build right after it builds this test library and
> then link to that library from my test program, the program executes
> without an exception!
>
> For some reason the library that the tests are building is working, and the
> libraries that are built using [ jam "-sTOOLS=mingw" stage ] are not
> working!
>
> I am very confused now. Any help will be greatly appreciated.
I was suggesting that it has to do with the way you compiled your
minimal.obj, and that you check that difference. However, since the
library built for testing works with your minimal.obj it must be a
difference in the way the library is being built for stage. So what
is the difference in command-line options with
bjam "-sTOOLS=mingw" stage
vs. whatever you invoked to build/run the boost regression tests?
Also, I guess you should consider whether there's just an outright bug
in the test program, although it looks too simple to be wrong.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk