|
|
Boost Testing : |
From: Beman Dawes (bdawes_at_[hidden])
Date: 2007-09-25 14:05:41
Sebastian Redl wrote:
> Beman Dawes wrote:
>> The patch has been applied, the revision number is present in at least
>> some of the .xml files, and "rev" is appearing in the column heads.
>>
>> But no actual revision number is appearing.
>>
> I underestimated the complexity of the stylesheets. Apparently, they
> generate a completely different tree from the original, and then process
> THAT.
> The revision number has to be explicitly transferred to the new
> stylesheet. The attached patch does that.
OK, applied. Cross your fingers:-)
> However, I cannot get the system to work for me at all, with or without
> the patch. It simply doesn't list any runners in the summary page, only
> the libraries. I have downloaded boostbook.zip, IBM_Canada_Ltd.zip, and
> the gcc, gcc-64 and intel Sandia results.
> There is, in fact, something very strange going on on my computer,
> possibly a bug in my xsltproc. An XSLT variable contains values it
> cannot possibly contain, or simply forgets its value.
>
> Either way, I cannot currently verify my changes myself.
While it would obviously be better to do some testing first, the patch
was so simple I was willing to take a chance.
What is motivating this is that I'm trying to clear various Boost.System
and Boost.Filesystem failures, and am finding it very tedious
determining if a fix was actually present for a given test run. The
testing system works pretty well for the most part. But some of the
rough edges make it very hard to use. So I'm hoping improvements like
showing the regression number and Noel's enabling of POSIX timeouts will
have a disproportionately large positive impact on usability.
Thanks,
--Beman