[Box Backup-dev] RC3 test failures on SLES9 ppc64

[Martin Ebourne]

On Mon, 2006-02-20 at 21:02 +0000, Chris Wilson wrote:
> I have no idea why, but the first time I ran configure, it detected 
> HAVE_VALID_DIRENT_D_TYPE. This is not true on this system: d_type is 
> always DT_UNKNOWN. This was causing the RaidFile tests to fail, because 
> they weren't finding any files.
> 
> Rerunning configure causes it to fail the test for 
> HAVE_VALID_DIRENT_D_TYPE, and all tests now pass.
> 
> Do you have any idea what could have caused that test to fail the first 
> time? Unfortunately, I don't have the configure logs any more, and I can't 
> reproduce the failure.

It actually runs the following code for this test:

          DIR* dir = opendir(".");
          struct dirent* res = NULL;
          if(dir) res = readdir(dir);
          return res ? (res->d_type==DT_UNKNOWN) : 1;

The only way that could succeed is if it successfully opendirs and
readdirs, and then d_type is not DT_UNKNOWN.

All I can think of is if d_type is uninitialised, but I don't think so.
Certainly the same code does the same thing on Fedora, and I've ran it a
hundred times and never seen that.

Unless of course it was a previously configured tree from somewhere else
and the value was cached. Not sure how/if that could happen.

Cheers,

Martin.