[Omake] FS case-sensitivity testing - if the name is non-alpha, should we still test.

[Aleksey Nogin]

On 04.06.2007 11:29, Jason Hickey wrote:

>> So the question is - should OMake be changed to only resort to
>> "invasive" testing (via the tmp file creation) when it actually cares?
>> The issue here is that the "audit" of access failures is becoming more
>> common (SELinux, etc, etc), so these tests are not always 100%
>> harmless (as they cause people to start wondering what is going on)...
> 
> Hmm, could be.  I don't quite understand your example, but in the end it
> is the equality we care about.  So, even if there is a numeric name
> along the path, we still care, for example, if the following are different.
> 
>     /usr/kde/3.5/bin/KONSOLE
>     /usr/kde/3.5/BIN/konsole

Note that currently OMake will test at every level. However, it does not
make sense to test at the /usr/kde level (since "lowercase" /usr/kde/3.5
and "uppercase" /usr/kde/3.5 are the same thing anyway), it is
sufficient to test the /usr/kde/3.5. Not coincidentally, the /usr/kde is
exactly the one where testing is especially problematic (as there may be
no existing alphabetical entries to test with).

> Also, we should probably consider using access(2) (or similar, based on
> stat) to detect the case where the directory is not writable.  This
> won't help if the directory looks writable but the fs is read-only, but
> I think we would still catch the vast majority of cases.

This also sounds like a good idea.

Aleksey