Re: Automated report: NetBSD-current/i386 build failure

["Simon J. Gerraty" <sjg%juniper.net@localhost>]

Andreas Gustafsson <gson%gson.org@localhost> wrote:

> [External Email. Be cautious of content]
> 
> 
> Martin pointed me to this error some 63 lines from the end of the log:
> 
>   --- dependall-tests ---
>   nbmake[7]: nbmake[7]: don't know how to make t_cabsl.cc. Stop
> 
> I think the reason I didn't find it myself is that I have developed a
> habit of searching for the message "Error code 1" (or similar with

I find "stopped" the key.

FWIW we do all our builds in meta mode, with a .ERROR target that copies
failing .meta file to a well known plase ($SB/error where SB=`realpath src/..`)
if such a file exits you *know* that was the cause of failure.
That only applies if a target was being built.

If a makefile caused the failure, you have to look for first instance of
"stopped" in the log to see why and where.
In that case we also spew a lot of info via MAKE_PRINT_VAR_ON_ERROR
which greatly aids triage (important with 2k developers)

> another number) which used to be printed by make, but that's no longer
> there.  Bracket also looks for that string as part of its heuristics
> for deciding how much of the build log to include in the email report,
> which is why this report didn't include any of it.
> 
> It would be helpful for both human and robotic users if error messages
> consistently included the word "error", or if there was some other easy
> way of identifying them in the build log.

The regex 'make.*stopped' is the best clue to look for since it will
always be present.