Re: [Lilypond-auto] Issue 3473 in lilypond: Patch: Rewrite STEPMAKE_GET_VERSION.

[lilypond]

Comment #18 on issue 3473 by address@hidden: Patch: Rewrite  
STEPMAKE_GET_VERSION.
http://code.google.com/p/lilypond/issues/detail?id=3473

> We don't have to take care of regressions in case everything is building  
> fine; it automatically means that there is *no* regression...

Uh, no?  People build with a variety of systems from source (is /bin/sh  
Bash?  Is it Dash?  Is it somewhat else?).  And the systems that are used  
for regtesting are not the same as the systems we use for crossbuilding.

The whole _point_ of the autoconf system is to ensure building on a  
_variety_ of different systems rather than a controlled subset.  So no,  
just because _we_ don't hit a problem does not at all mean that  
everything's fine.

We have punted for Make and require GNU Make.  But our configure scripts  
will accept whatever the host considers /bin/sh, so this is really an area  
where it's pretty easy to get ugly surprises.

If we have ugly quoting problems, then wrapping stuff in shell functions is  
no guarantee that the quoting problems will be covered.  Admittedly, eval  
is a source of trouble, so being able to avoid that tends to be prudent.

At any rate, changes to the autoconf system have to count as being largely  
untestable since our tests only cover a tiny fraction of what the system is  
supposed to cater for.  Also any problem we uncover in the cross-building  
phase is expensive, and hard to track down.

--
You received this message because this project is configured to send all  
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings