Re: [Bug-gnulib] Re: CVS Bug? or User error?

[Alexandre Duret-Lutz]

| Index: lib/am/yacc.am
| ===================================================================
| RCS file: /cvs/automake/automake/lib/am/yacc.am,v
| retrieving revision 1.16
| diff -u -p -r1.16 yacc.am
| --- lib/am/yacc.am    18 Apr 2003 08:54:34 -0000      1.16
| +++ lib/am/yacc.am    1 Jun 2004 13:46:41 -0000
| @@ -17,6 +17,7 @@
|  ## 02111-1307, USA.
|  
|  ?GENERIC?%EXT%%DERIVED-EXT%:
| +if MAINTAINER_MODE
|  ?!GENERIC?%OBJ%: %SOURCE%
|  if %?MORE-THAN-ONE%
|  ?GENERIC?    $(SHELL) $(YLWRAP) %SOURCE% y.tab.c %OBJ% y.tab.h %BASE%.h 
y.output %BASE%.output -- %COMPILE%
| @@ -48,3 +49,7 @@ else !%?MORE-THAN-ONE%
|       sed '/^#/ s|y\.tab\.c|%OBJ%|' y.tab.c >%OBJ%t && mv %OBJ%t %OBJ%
|       rm -f y.tab.c
|  endif !%?MORE-THAN-ONE%
| +else !MAINTAINER_MODE
| +     ## Skip this target unless in maintainer mode.
| +     @:
| +endif !MAINTAINER_MODE

Hi Derek!

I'm not opposed to such change, but as I have little time these
days I'm opposed to incomplete patches :)

Since you change the way MAINTAINER_MODE and yacc rules
interact, you should update both documentation and NEWS it.  I'd
also like a proof that it works.  Preferably a test case that
would fail without your change.  The above patch is wrong
because one of the two possible first line of the rule is inside
the conditional (%OBJ%: %SOURCE%), and the other is outside
(%EXT%%DERIVED-EXT%:).  Also in the !MAINTAINER_MODE case it
would be nice to justify why need you keep a no-op rule instead
of removing the rule entirely.  Finally, if yacc.am is changed,
then lex.am should probably be changed for consistency.

Hope I haven't frightened you :)
-- 
Alexandre Duret-Lutz