Re: 2.59d on OpenServer

[Ralf Wildenhues]

* Kean Johnston wrote on Sun, Jun 18, 2006 at 06:13:10PM CEST:
[...]
> and simply changing from:
> 
>   ac_subst_vars="word1 word2 word3"
> 
> to
> 
>   ac_subst_vars="word1"
>   ac_subst_vars="$ac_subst_vars word2"
>   ac_subst_vars="$ac_subst_vars word3"
> 
> fixed the problem. The actual line length limitation *appeared*
> to be the contents of the quoted string, *before* expansion.
> But I have just tried to create a small isolated test case
> that simply has:
> 
>   #!/bin/sh
>   LONGVAR="WORD WORD WORD ..."
>   echo "$LONGVAR"
> 
> for an extremely long list of "WORD WORD ...." (about 70K)
> and its working just fine.

Did you put anything else in the test script?  My impressions from
http://lists.gnu.org/archive/html/autoconf/2006-05/msg00141.html
was that some memory corruption occurs, not always manifesting itself
in a hard error at the time of corruption, so you might want to paste
a configure script after your test.  ;-)

> So the limitation isn't immediately
> obvious. Something about the way a configure script is run, or
> what it does at startup or something else is what is causing
> the limitation to be reached.

Try /usr/bin/posix/sh, and use BIN_SH=xpg4; export BIN_SH.
This is one major difference between Autoconf 2.59 and current.

> I will play around with this some more and try to get to the
> bottom of it, because that may help lead me to where to look
> in teh shell to avoid this ridiculous limitation. Mind you,
> assigning 12K of text in a single assignment is a bit on
> the ridiculous side too :)

No, arbitrary limitations (other than outside ones such as available
main memory) in a general purpose tool are ridiculous.  Assigning as
much as possible in one go is the most efficient way with all modern
shells.  But wouldn't Autoconf be boring if all shells were that way?
;-)

> But my gut feeling, and my vague recollections of the problem
> lead me to believe that the limitation is before expansion,
> not after.

Cheers,
Ralf