Subject: Re: [Boost-build] Incomplete builds with generated header + -jN
From: Eric Woodruff (eric.woodruff_at_[hidden])
Date: 2009-09-09 16:11:17
Is there an open bug on this? I'm reproducing a similar problem of
incomplete builds using the bjam and boost-build from boost 1.40.
I have a validation script that copies my sources to a new sandbox,
performs a build and then performs a -n build to make sure the previous
build was complete. There are around 35,000 targets (with all of the
boost headers, etc), 360+ need to be built from a clean state and about
39 of them are left unbuilt after the first pass. The validation does
succeed with -j1.
Jurko GospodnetiÄ wrote:
> Hi Mark.
>>>> I suspect you might either have to produce a self-contained example,
>>>> or debug this yourself, like by adding prints in make*.c that explain
>>>> in more detail what happens with the file that is not rebuilt.
>>> Yes, I thought that would be the case :-(. I'll keep digging.....
>> I did keep digging for quite a while, but didn't manage to get to the
>> bottom of it. In the end I worked round the problem by adding code to
>> compare the value of counts->updating in make.c with the value of
>> counts->made in make1.c, and if there's an underbuild then spawn jam
>> again with the same arguments. It's pretty nasty, but it works.
> Any chance of you producing a small self-contained example
> reproducing the problem? And possibly post it as a defect report/feature
> request on http://zigzag.cs.msu.su/boost.build so it does not get lost?
> Best regards,
> Jurko GospodnetiÄ
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build
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