On Thu, Jan 15, 2009 at 9:05 PM, Ken Williams <[email protected]> wrote:

> I'm not sure what we're talking about.  Is it "1 while unlink $f"?
>

No.

There are a couple places that use rmtree to delete stuff.  Base.pm and also
some of the t/lib files.

I've found that if I load my windows machine enough, I can cause rmtree to
fail due to permissions -- usually because a file couldn't be deleted.  Best
hypothesis is that it's due a virus scanner that is too slow due to the
loading to check the file and get out of the way for deletion.

The result is that various tests that check "! -e $foo" then fail.

Note -- to "replicate" this bug today, I was running the following:

* Lotus Notes
* Ubuntu in a Virtualbox VM  -- running "visitcpan" on a drive shared from
the Windows host
* Firefox
* SSH window
* (plus background corporate network update/backup stuff)

This machine was so heavily loaded I was getting mouse and keyboard lag.
;-)

I've seen the error on less loaded machines, but my earlier attempts to
recreate the problem weren't working so I tried to max out the system and
see if that was enough to break things.

Sadly, even my proposed "for (1 .. 20) " type loop around rmtree were
insufficient to avoid the problem.  So it might take more effort to figure
out where the virus scanner is checking the file -- possibly during rmtree?
But I'm rapidly getting beyond (a) my skills to debug and (b) my interest in
fixing the problem.

I'm open to suggestions.

-- David

Reply via email to