Re: make-3.81rc1 mingw not failing on 'g++ -M' error

[Matt England]

At 3/7/2006 01:58 PM, Eli Zaretskii wrote:
> > Date: Tue, 07 Mar 2006 13:11:31 -0600
> > From: Matt England <address@hidden>
> > Cc:
> >
> > In a different part of my real makefile (which is much more complex the 
> one
> > in this thread), make will continue to run some of the make/compile 
> jobs in
> > the background, asynchronously after experience an error, even though I
> > wasn't running 'make -j'.
>
> Sorry, I looked at your transcript, and I cannot make heads or tails
> out of it.  Can you explain how do you see that ``jobs are run
> asynchronously in the background after an error'', and then show in
> the transcript the evidence of this asynchronous operation?

If I could do that easily, I would.  I tried massively to do that in the 
last output.  I'm sorry I won't be able to help further.

If I could reference the archive post to prove it, I would...but there's a 
delay, apparently, in the archived email showing up at:
<http://lists.gnu.org/archive/html/make-w32/2006-03/threads.html>

> Is it possible that some or all of your rules' commands use the same
> technique as I commented on in another message: i.e. that they issue
> multiple commands as a single shell command line?  Could this explain
> what you see?

First answer: Most of my rules are single-command-line rules.  The one 
previously was an anomaly; further, when I turned it into the single-line 
rule, the same problem happened.  I'll show that further in a later email.

Second answer: I expect the multi-line 'set -e' shell command to work just 
as well as the single-line shell command.

> > If someone wants my big, hairy set of Makefiles, they are welcome to
> > it...but they are sizable.  As much as I'd like to boil this down to a
> > "minimal" Makefile to give to this community, this could be a significant
> > effort to reproduce this problem in another Makefile.
>
> Please show at least the parts that generate the commands you show in
> the transcript.  They seem all to come from a single rule, perhaps a
> pattern rule or some such.  It's impossible to debug this without
> seeing at least that much.

...and it's really difficult for me to narrow down the Makefile.  It's a 
massively-$(eval)-based system, with a "core" makefile and 
application-specific ones; call it my flavor of cmake/qmake, but with more 
power (imho).

> > Finally the make job stopped on another error.  :)
>
> But I didn't see any other errors, this one was the first!

They are there, please look closely at the output.

The asynchronous stuff just runs in the background and appears on the 
screen.  To the untrained eye, it looks like synchronous output.  But it's 
not.  It's rather confounding

I realize this may be a shell problem.  If y'all could tell me *exactly* 
which mingw/msys configuration you are using to run these shells, I might 
be able to tag along.

I suspect that some of these problems are that mingw/msys may just be 
unreliable and/or inconsistent...but I'm speculating here.

In the end, I suspect I'm going to need to mingw-cross-compile (on Debian, 
for example) in order to make reliable, automated builds with the same 
robustness that I get on my unix/linux systems.

-Matt