Charles Lane <[EMAIL PROTECTED]> wrote:
|
| Charles Bailey [EMAIL PROTECTED] wrote:
| > We could make a case for forcing GLOB_NOCASE always under VMS. It does remove
| > one of the documented options, but since readdir will always return lower
| > case filespecs, it's not useful, and it does create protability problems
| > as with this test.
|
| The question is: if you're going to override default behavior, do you
| want the module to decide that you *really must have* this particular
| flag? I'm not sure...
As a general principle, it makes me leery, but I'm not sure what sense
~GLOB_NOCASE makes under VMS, since the filesystem will always return
all-lowercase filespecs. Put a bit differently, readdir() (well, Files-11,
but readdir picks which case) already overrides the cshish default behavior;
maybe it'd be better to propagate this into the execution of File::Glob,
favoring practical effectiveness over fidelity to documented behavior
(at least within the XS code).
| BTW, did you have a chance to look at the tainting of directories in
| mg.c ? (message <[EMAIL PROTECTED]>)
| The directory protection that triggers tainting seems wrong
| to me ... to the extent that it would be a security hole...but
| someone more familiar with tainting should take a look.
We really should fix the spelling of that bit for 5.6.1. In practice,
it works out fine, since vando_by_name() maps S_IW* to the same ARM$M_WRITE
bit, so S_IWUSR has the same force as S_IWOTH.
Regards,
Charles Bailey [EMAIL PROTECTED]
