RE: [jamboost] Re: bjam quits on failure to found a target?

From: Aleksey Gurtovoy (alexy_at_[hidden])
Date: 2003-05-07 03:19:06

David Abrahams wrote:
> >> Sure. You can grep for the error message in tools/build/*.jam and
> >> replace 
> >> 
> >> EXIT
> >> 
> >> with
> >> 
> >> ECHO warning: 
> >
> > Won't that break the standalone builds? 
> 
> How so?

I meant "hurt usability of the standalone builds".

> 
> > I mean, EXIT is probably quite appropriate when you are just
> > building the library itself; it's just not appropriate in the
> > context of the batch regression testing.
> 
> I'm not picky; BBv1 is a dead end anyway. We're just keeping it
> limping along until BBv2 gets finalized (soon).

OK, if you say so. I've replaced EXIT with ECHO, but unfortunately it 
didn't move us very far forward - now bjam quits with a run-time 
error in 'find-compatible-subvariant' rule - please see the attached 
log. The full one is available here -
http://www.meta-comm.com/engineering/resources/bjam_log.zip.

So, even we had BBv2 on hands, it wouldn't really solve the problem of 
one erroneous library breaking the whole regression process, since 
explicit EXIT is not the only way it can happen (bjam crashing aside). 
And even if we took precautions for that (e.g. by having bjam handle 
each library separately), one error somewhere in 'boost-base.jam' is 
still all it would take to bring the whole thing down again.

But then, given that it won't happen that often, it might be 
acceptable. What is not acceptable IMO is that currently there is no
way for anybody besides the person who runs the tests to find out 
whether the regressions are up and running or totally broken. Not to
mention that the runner has to explicitly report the failure to the
list, post logs, etc. 

I propose we modify 'regression-logs.pl' script to handle, in 
addition to the current .html-s, bjam log uploads - linking them from
the page, and making the table row painfully red if there is a log, but
no compilers in html-s (or whatever heuristic we come up with to detect
the cases like the one we are dealing right now).

Aleksey

 ------_=_NextPart_000_01C31471.53937AE0 Content-Type: application/octet-stream;
name="bjam_07_may_03_truncated.log"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment;
filename="bjam_07_may_03_truncated.log"

[Attachment content not displayed.] ------_=_NextPart_000_01C31471.53937AE0--