From: Curt Arnold <carnold@ho...> - 2004-04-02 17:16:31
My guess is the FreeBSD installation has a link command on the path and
the Linux installation doesn't. The test checks the getIdentifier()
method which in this case would execute a "link /bogus" to get the
identification string from the Microsoft linker and return "link" if
there isn't one. The test checks for "Microsoft" or "link". If "link
/bogus" returns something else, then it would fail the test in that
On Apr 2, 2004, at 1:53 AM, Stefan Bodewig wrote:
> we've just recently installed Gump on a FreeBSD machine and one of the
> things that failed here but work on Linux is the testsuite for
From: Stefan Bodewig <stefan.bodewig@fr...> - 2004-04-05 06:43:56
On Fri, 2 Apr 2004, Curt Arnold <carnold@...> wrote:
> My guess is the FreeBSD installation has a link command on the path
> and the Linux installation doesn't.
Yes, FreeBSD seems to have a /bin/link which is essentially ln:
-bash-2.05b$ link /bogus
usage: ln [-fhinsv] file1 file2
ln [-fhinsv] file ... directory
link file1 file2
-bash-2.05b$ which link
-bash-2.05b$ ls -l /bin/link
-r-xr-xr-x 2 root wheel 52996 Mar 11 21:42 /bin/link
So the test needs to check whether it is running on Windows?
From: Curt Arnold <curt.arnold@ma...> - 2004-04-05 17:24:27
It is possible that you could be running Microsoft's tool chain using
WINE, so I'd hate to inhibit the MSVC toolchain on non-Windows
platforms. The test could be modified so that the output of the
FreeBSD "link" is considered an acceptable response for the test.
From: Stefan Bodewig <stefan.bodewig@fr...> - 2004-04-06 05:49:16
On Mon, 5 Apr 2004, Curt Arnold <curt.arnold@...> wrote:
> It is possible that you could be running Microsoft's tool chain
> using WINE, so I'd hate to inhibit the MSVC toolchain on non-Windows
Fait enough. Note that I've just modified the test to only run on
Windows, just the test, no application code.
> The test could be modified so that the output of the FreeBSD "link"
> is considered an acceptable response for the test.
After thinking about this test once more, I'm pretty sure that it isnt
that valuable at all. I pretty much doubt that MSVC is the only suite
around that ships with a link.exe and having the task adapt to the
output of each "link" seems as good as dropping the test completely.
The test would certainly fail on OpenVMS as well.