Tue Mar 9 20:31:46 PST 2004
John Dennis writes:
> With respect to bonsai, I just want to give a heads up that I'm working
> on it, as well as LXI, no point in duplicating efforts, additional help
> always welcome if you're expert.
> Well, I haven't had a chance to check the logs but its a standard build
> log, if the build failed there is likely a compiler error which should
> have an error message. Of course other things can go wrong, CVS errors,
> or even an error in the Tinderbox script. When Jim and I worked on it we
> were pretty careful to catch and report all errors, but I don't think we
> were clever enough to make sure all the errors messages started with the
> string "ERROR" making it easier to find, we'll fix that.
The problem seemed to be that we were getting false positives, ie.
error flags from clean builds.
Finding errors in builds seems to be possible when looking for
'***', however I'm not sure if this happens all the time.
> There is one other possibility, if you go to the bottom of the log and
> the log says the build was successful but Tinderbox does not agree (one
> case where you would search for an error and not find it) then we need
> to modify how the Tinderbox script checks for a successful build. What
> we do now is check the process exit status on every command we run. Make
> should be reporting the correct error status, but I wouldn't fall over
> in surprise is there is a broken make program out there.
Right. That seems to be the problem :-( In fact I did not find errors
listed in the summaries either. I would have thought that a message
describing during which step the error occured whould appear there,
> Or "make Everything" as was suggested by Michel. O.K. you can tell how
> long I've been around :-) Years ago the top level makefile didn't have
> nice targets and you had to do things manually, old habits die hard, the
> other suggestions are better.
Right. I those days building with Imake was really a nightmare.
This is a myth that has survived until today.
> > It should clobber every time. At least now.
> > And it should be done if a previous build broke.
> I'm not sure about clean clobber everytime, but its worth considering if
> others agree. I'm also not sure if a clobber is justified if the
> previous build failed. If there was a trival syntax error in one file,
> do you really want to blow everything away and wait a long time to see
> if its fixed? Wouldn't you rather start from where you left off and
> immediately verify the fix? At the moment the tinder client does not
> retain state from previous builds which would be necessary to implement
> this behavior, I can look and see if its possible.
'make Everything' would be just fine I think.
More information about the release-wranglers