On Mon, 29 Aug 2005, Scott Davis wrote:

> Might I ask what I can do to investigate this failure? As I delve into the
> "tests" subdirectory (and its subdirs) I can't really find anything
> interesting. The only thing I see is the xml file used for test 56 which
> tells me what was being attempted, but I don't see any log output that
> indicates the cause of the failure.

First, I seriously doubt that this test failure actually is a serious and true
failure in the curl code. Most test case failures that happen like this are
due to flaws and shortcomings in the test servers or test scripts.

That said, tou can do this to investigate and debug:

cd into the 'tests' dir and run only test 56 like this:

$ ./runtests.pl 56

If it still fails, it leaves its output files in the log/ dir (and it it
doesn't fail you can force it to leave those files there by using runtests.ok
-k)

If you can repeat a failure and you want to plunge into the inner workings of
curl and libcurl, you are better off first building curl static and with debug
symbols (configure --enable-debug --disable-shared) and then if you have gdb
enable you can run curl with gdb on that test case like this:

$ ./runtests.pl -g 56

This fires up gdb with all necesary bits setup properly so you can set a
break-point or two and then type 'run' to actually perform the test.

If this doesn't cover most of it, ask again!

Oh, and the test case file format is described in the tests/FILEFORMAT file
should you wonder what all the various parts mean.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html