Re: [boost] [thread] tss_hooks.cpp "leak" fix in 1.34?

From: Roland Schwarz (roland.schwarz_at_[hidden])
Date: 2006-09-17 12:02:02

Peter Dimov wrote:
> Jeff Garland wrote:
>> Peter Dimov wrote:
>>> We have another complaint about the "leak" in Boost.Threads:
>>>
>>> http://lists.boost.org/boost-users/2006/09/22356.php
>>> http://lists.boost.org/boost-users/2006/09/22358.php
>>>
>>> Should we merge the new version of tss_hooks.cpp into the 1.34
>>> branch?
>> Based on the frequency of reporting and the 'apparent seriousness' I
>> think it should be pulled in.  It also seems to be a pretty low risk
>> change.  Here's a quick link to the diff for Thomas ;-)
>>
>> http://boost.cvs.sourceforge.net/boost/boost/libs/thread/src/tss_hooks.cpp?view=diff&r1=text&tr1=1.5&r2=text&tr2=1.4&diff_format=h
> 
> Nope,
> 
> http://boost.cvs.sourceforge.net/boost/boost/libs/thread/src/tss_hooks.cpp?view=diff&r1=text&tr1=1.5&r2=text&tr2=1.3&diff_format=h
> 
> I'm afraid. :-)

While having only looked into this very cursory, I cannot imagine that 
calling TLSFree will have anything to do with the reported leak.

This AFAIK only frees a system resource, which is not covered by the 
reporting tools. (But I might be wrong.)

The reported leak also semms to have nothing to do with tss at all.

Besides this it is known that there is an annoying report of TSS leaks,
which is known to be harmless. The problem with this is (as far as I 
remeber) that in fact there is no leak, altough the tools are reporting
one, which has to do with the way where memory is released.

But to be honest, I have not looked into the code for some time now.
Last time I tried to get rid of this problem while I was writing the
code that allowed the static linking of the TSS lib.

It would be very helpful to get an exact report in which version
the leak is occuring, and also an example using TSS which shows the leak.

Roland