On 03 January 2007 01:05, Paul Smith wrote:
> On Tue, 2007-01-02 at 20:35 +0000, Dave Korn wrote:
>> On 02 January 2007 20:16, Hiebert, Darren (MS) wrote:
>
>>> Under what circumstances can a file be reported by make as not existing
>>> when it does exist?
>
>> When it doesn't exist in the directory(ies) that make is searching in for
>> it.
>
> Another possibility is you're running afoul of GNU make's directory
> caching. To improve performance GNU make will cache the contents of
> directories the first time it needs to read them. Thereafter make's
> internal cache is updated by make itself, as it updates targets.
>
> However, if you have sneaky rules which create files that aren't listed
> as targets in your makefile, then make has no way to know that they've
> been created and so its internal cache will not reflect the reality on
> the filesystem.
>
> This can cause GNU make to think that files don't exist which actually
> have been created.
Say, I may have just run into something similar. Is this directory cache
used when evaluating $(wildcard ...) by any chance? And is it perhaps not
flushed after recursively invoking a sub-make?
cheers,
DaveK
--
Can't think of a witty .sigline today....
_______________________________________________
Help-make mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-make
