From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-07-11 00:17:23
Beman Dawes wrote:
> At 02:43 PM 7/10/2003, klapshin wrote:
> >--- In jamboost_at_[hidden], Beman Dawes <bdawes_at_a...> wrote:
> >> A few of the tests worked fully. But for many others, the only commands
> >> given were mkdir's. No other sign of actions in those directories.
> >Yep, that's exactly the behavior I've seen. Try to update bjam again
> >from CVS. I belive yesterday Vladimir fixed some bugs; the new version
> >works fine for me. Don't know if the "-a" bug is gone, or just better
> >hidden now. :)
> OK, I've done a CVS update and then rebuilt bjam. -a seems to be working
> again, although it is hard to tell if it skipped one or two actions.
It's quite easy actually. Before doing anything, bjam prints
...updating XXX target...
in the end it should print
...updated YYY targets....
If YYY is less then XXX and there were no skipped targets, we're in problem. I
think I'd rather explain the bug I've introduced last week and fixed
Whenever each target is built, all dependents of this target are notified, and
when all outstanding dependencies of a given target sent notification, the
target itself is built. The current version of bjam has certain Perforce
change merged in --- and each target with #includes have another "internal"
target which represent included targets. The problem was that when "internal"
target was built, the code was failing to send correct notification. As the
result, dependents were never considered ready.
And when "-a" was present, *all* internal nodes were marked for updating, and
consequently, nothing was ever considered ready for updating.
I'm hoping that's the last bug in my changes of dependency scanning.
> To unsubscribe from this group, send an email to:
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
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