This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: YA call for snapshot testing

On Sun, Jan 22, 2012 at 08:11:18PM +0100, marco atzeri wrote:
>On 1/22/2012 5:57 PM, Christopher Faylor wrote:
>> On Sun, Jan 22, 2012 at 08:30:23AM +0100, marco atzeri wrote:
>>> I saw the problem in erratic way also before running updatedb.
>>> But it is very evanescent and usually linked on how find is called.
>> Could you try the latest snapshot?
>snapshots 20120122 07:28:45 seems to have solved the issue

Looks like my guess was correct.  I added refence counting to Cygwin's
inaptly-named fhandler structures recently to work around a problem
where a signal handler closed an fd while it was in the middle of a
read.  This caused the memory associated with an fd to be deleted only
when the last thing referencing it was done with it.

The very odd thing was that my implementation seemed to work right from
the beginning.  That has made me wonder what I got wrong.  What I got
wrong was the dup*() family of functions.  If you dup a fd its reference
counter was not reset so the memory associated with it was not deleted.

Thanks for confirming.


Problem reports:
Unsubscribe info:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]