Boost logo

Boost-Build :

Subject: Re: [Boost-build] tools/build/v2 tests causing release branch reporting to fail
From: Beman Dawes (bdawes_at_[hidden])
Date: 2009-01-12 17:27:30


Steven Watanabe wrote:
> AMDG
>
> Beman Dawes wrote:
>> Noel Belcourt has finally tracked down why the branches/release
>> regression reporting is failing, and it is because the Boost build
>> test results from the test runners contain invalid XML. This in turn
>> causes the report generation to fail.
>>
>> Noel reports getting an error from all release results.
>>
>> # target doesn't exists, building
>> # Merging "/incoming/branches/release/
>> processed/Sandia-darwin-ppc.xml" with expected results...
>> # Skipping
>> "/tmp/kbelco/reporting_release/release/incoming/branches/release/processed/Sandia-darwin-ppc.xml"
>> due to errors
>> (/tmp/kbelco/reporting_release/release/incoming/branches/release/processed/Sandia-darwin-ppc.xml:97765:19:
>> not well-formed (invalid token))
>>
>> The errors are of this form:
>>
>> failure {{{
>> "['bjam',
>> '-sBOOST_BUILD_PATH="/Volumes/Application/scratch/kbelco/boost/tools/build/v2/test"',
>> '-d0', '--quiet', 'toolset=darwin-4.0.1',
>> '--test-config="/Volumes/Application/scratch/kbelco/boost/tools/build/v2/test/test-config.jam"',
>> '--ignore-toolset-requirements']" returned 1
>> }}}
>> reason {{{
>> error returned by bjam
>> }}}
>> STDOUT {{{
>> Generating a CPP file...
>> ...skipped <pOther/bin/darwin-4.0.1/debug>other-obj.o for lack of
>> <pOther/bin/darwin-4.0.1/debug-object(c-scanner)@107>other-obj.cpp...
>> ...failed updating 1 target...
>>
>> }}}
>>
>> I've confirmed his results, and further find that the reporting runs
>> OK if the invalid XML text is removed.
>>
>> Perhaps something didn't get merged to release branch when the build
>> tests were added to trunk?
>
> Right. The last change to release Boost.Build was on November 9th.
> The problem is that the release branch Boost.Build does not know
> how to generate the proper xml output.

Volodya made most of the changes since then, but he isn't the only one
who hasn't merged changes. We really need to smooth out the merge
procedures for build changes. I did it once and got my wrist slapped
since some weren't really ready for prime time. Only the folks who make
the changes to trunk know if a given change should also go into release.

Regardless, whichever changes that are needed for 1.38.0 need to be
merged right away so we can regain lost momentum.

Thanks,

--Beman

--Beman


Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk