Re: PATCH: Fix leak during Y shortcircuit glob qualifier

[Mikael Magnusson <mikachu@xxxxxxxxx>]

On Thu, Sep 24, 2015 at 7:18 AM, Bart Schaefer
<schaefer@xxxxxxxxxxxxxxxx> wrote:
> On Sep 24,  5:38am, Mikael Magnusson wrote:
> }
> } % : ./*(-.Y1N); ls -l /proc/$$/fd
> }
> } was my test case, so not much would happen to $PWD in that particular
> } case. I tried */*/*(-.Y5N) instead now, and nothing untoward seemed to
> } happen then either (without and with my patch, so no weird pwd
> } changing or errors about double frees, respectively).
>
> Glancing at scanner() in glob.c and lchdir() in utils.c, it appears that
> the working directory would only be changed if the full path to the target
> directory is VERY long, i.e., does not fit in a PATH_MAX buffer.  So it
> probably is necessary to restore, but also probably really difficult to
> construct an example actually goes through that code path.

I've gone the safe route for now and only committed the closedir(lock) part.

-- 
Mikael Magnusson