[Omake] SVN Commit: OMake Build System [0.9.8.x] (Rev. 9376)

[Aleksey Nogin]

On 23.06.2006 12:59, Jason Hickey wrote:

>       Added the bounds-check.  Implicit calculations are not guaranteed
>       to terminate in general, but simple cycles will be rejected.
>       A simple cycle is one where a file depends on itself (through
>       some chain of dependencies).

This fix is definitely better than nothing (it will only kill those 
cases that would have resulted in an infinite loop before), but it's not 
ideal - it's interaction with a cache may result in a "missed opportunity".

$ cat OMakefile
.DEFAULT: file.a

%.e: %.c
     cp $< $@

%.d:
     touch $@

%.b: %.d
     cp $< $@

%.c: %.b
     cp $< $@

%.b: %.c
     cp $< $@

%.a: %.b %.e
     cp $< $@

$ omake
*** omake: reading OMakefiles
*** omake: finished reading OMakefiles (0.1 sec)
*** omake: 7/8 targets are up to date
*** omake: failed (0.1 sec, 0/0 scans, 0/0 rules, 8/33 digests)
*** omake error:
    File OMakefile: line 1, characters 0-16
    Do not know how to build "file.a" required for "<phony <.DEFAULT>>"
Exit 123

But the file is actually buldable - and if the "%.b: %.c" rule is 
deleted, then OMake does not have any problems figuring it out.

The problem here that file.c is examined first as a potential way of 
building file.b. Since the only way to build file.c is to use file.b, we 
detect a loop and mark file.c as unbuildable. But this is not true - it 
is only unbuildable if we do not have _another_ way of building file.b, 
not involving file.c, and here we do have such a way.

I am not sure whether it's worth fixing this. But I though that it is 
worth pointing out this issue.

-- 
Aleksey Nogin

Home Page: http://nogin.org/
E-Mail: nogin at cs.caltech.edu (office), aleksey at nogin.org (personal)
Office: Moore 04, tel: (626) 395-2200