RE: miniserv.pl hangs on module update

[Curtis D. <Cu...@Gr...>]

>> >> I seem to have recently acquired a bug where I cannot do a remote url
>> >> install of a module.
>> >>
>> >> So anyway, miniserv.pl is still hanging. Endless hourglass when
>> >I attempt to
>> >> install a module from local or remote source. It looks like
>> >miniserv.pl is
>> >> spinning and enlessly spawning rm, over and over. Any ideas on
>> >how to debug?
>> >
>> >Odd .. what module were you trying to remotely install?
>>
>> I've figured it out. This problem stems from an earlier patch I
>submitted to
>> Jamie that overloads stat with objects via the File::stat
>module. (I prefer
>> $foo->mtime to $foo[4].) Anyways, what happened was inside
>&tempname() where
>> it never exited the while(1) loop because the array was actually
>an object
>> an thus it was never seeing the permissions it expected to see.
>>
>> I've updated the patch to use File::stat without overloading, and the
>> problem went away. That module is part of the standard perl
>distribution so
>> I can only hope Jamie will consider it.
>
>I'm a bit reluctant to use perl modules when there is already a built-in
>function for doing the same thing, in order to cut down on load time and
>to make it easier for people to run webmin on stripped-down
>versions of perl.

I understand. Performance is very important. My preference is always for
code that is readable and self-documenting. That said, the difference
between $inode[9] and $inode->mtime isn't that big. B'sides, he who knows
what an hell an inode and an mtime is, probably already learned the numbered
elements in stat(). ;)

>> However &tmpname() still needed some serious help, so I rewrote
>it. Some of
>> it is shamelessly swiped from the FAQ. ;) Attached is the patch.
>
>Your patch basically does the same thing as the existing
>tempname() function,
>just using File::stat right (apart from removing all my hard-coded
>uses of /tmp :-)

It achieves the same purpose. And the patch is cumulative. Addtional changes
are:
	- will get TMP from the environment if not /tmp (cross-platform futures?)
	- system call now explicit, doesn't use PATH (security?)
	- bail out counter in until() instead of while(1) loop (safety?)
	- enhancement to make_date() I sent you awhile back
	- further extension to our module versioning idea
	- extra security on diffs we talked about previously

I think that's the jist. But I assuume you already caught those from the
source and my comments therein.

../C