Re: How to pinpoint aclocal failure

[Yang Tse]

2008/8/3, Eric Blake wrote:

> [Please keep the list in the loop]

[Sorry for that. I usually simply hit 'reply' and forget to double
check if the mailing list address is actually in the 'from' or 'reply
to' field of original message.]

> So you are saying that the presence of the file containing the definition
> of MACRO_A, even though MACRO_A is not invoked, is the trigger for m4's
> success or failure?  Can you post the file containing MACRO_A for review?

Exactly, yes.

But, since I've written that, there has been some progress, and
finally the problem seems to be an OS issue related with exhausted
memory while there is still plenty of swap space available. Deeper
investigation will be done but no further info is available at this
time. At least now we can dig in the right p

So finally it seems that all the auto* tools and our own macro files
can be ruled out of the issue, at least in a first instance.

Since this is certainly related to the memory utilization I have one
question directly related with the ordering of macros in a given file.

Given the following macro file
http://cool.haxx.se/cvs.cgi/curl/m4/curl-reentrant.m4?rev=1.1&content-type=text/vnd.viewcvs-markup
The macro named CURL_CONFIGURE_REENTRANT (what I've been calling
MACRO_A) )is declared at the end of the file and uses macros which are
previously declared in the same file.

Is there any specific ordering in the macros inside this file that
imposes a higher or lower memory utilization for the autoconf tools ?

As you can see, right now what is being used inside that file is
'define before use', but I don't actually know if 'use before
defining' would be a better choice, due to inverse processing or
something else.

Which ordering imposes less work or less memory requirements on
autoconf and autom4te ? And, if you happen to know it,  on automake
and aclocal ?

-- 
-=[Yang]=-