Re: pattern rules with absolute paths and chaining (Was: implicit rules with absolute paths and chaining (Was: targets starting with '/'))

[Noel Yap]

I haven't tried the patch out yet, but the explanation doesn't seem to explain 
why replacing the absolute paths with relative paths fixes the problem.

Noel

Paul D. Smith wrote:

> There is a known issue with GNU make 3.80 where any pattern that refers
> to a non-existent directory is dropped.  I'm not sure what the rationale
> behind that is, besides efficiency.
>
> But any makefile that relies on real make rules to create directories,
> rather than using the $(shell ...) hack, will need this fixed.  It's
> been fixed in CVS already.
>
> I attach the patch below.
>
>
>
>
> ------------------------------------------------------------------------
>
> Index: rule.c
> ===================================================================
> RCS file: /cvsroot/make/make/rule.c,v
> retrieving revision 1.35
> retrieving revision 1.36
> diff -u -B -b -r1.35 -r1.36
> --- rule.c      24 Feb 2004 13:50:22 -0000      1.35
> +++ rule.c      22 Mar 2004 15:11:49 -0000      1.36
> @@ -135,20 +135,6 @@
>                  nonexistent subdirectory.  */
>  
>  	      dep->changed = !dir_file_exists_p (name, "");
> -#ifdef VMS
> -              if (dep->changed && strchr (name, ':') != 0)
> -#else
> -             if (dep->changed && *name == '/')
> -#endif
> -               {
> -                 /* The name is absolute and the directory does not exist.
> -                    This rule can never possibly match, since this dependency
> -                    can never possibly exist.  So just remove the rule from
> -                    the list.  */
> -                 freerule (rule, lastrule);
> -                 --num_pattern_rules;
> -                 goto end_main_loop;
> -               }
>             }
>           else
>             /* This dependency does not reside in a subdirectory.  */
> @@ -159,7 +145,6 @@
>         max_pattern_deps = ndeps;
>  
>        lastrule = rule;
> -    end_main_loop:
>        rule = next;
>      }
>  
>
>
> ------------------------------------------------------------------------