X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
From: CBFalconer <cbfalconer AT yahoo DOT com>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: libstdc++ writev/2.04/patches upstream?
Date: Sat, 19 Jul 2008 18:55:33 -0400
Organization: Ched Research http://cbfalconer.home.att.net
Lines: 42
Message-ID: <488270E5.84BCCBF5@yahoo.com>
References: <200807070405 DOT m67451dZ010910 AT delorie DOT com> <BAY139-W19808358F507B7930E7F6EE6940 AT phx DOT gbl> <200807071920 DOT m67JKA4v032518 AT envy DOT delorie DOT com> <487290E2 DOT 86197C00 AT yahoo DOT com> <200807072300 DOT m67N0t0Y005696 AT envy DOT delorie DOT com> <4872DFDB DOT 3E4BE7C1 AT yahoo DOT com> <200807080501 DOT m6851VZ8015818 AT envy DOT delorie DOT com> <4873E90E DOT B2B957F9 AT yahoo DOT com> <48742FD3 DOT 4080701 AT iki DOT fi>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Trace: feeder.motzarella.org U2FsdGVkX18yNIpTpWEi7abl6zXQuJp/mvbjDo+j//Kims/hgzQy6WBIoKIiUJS1W8GSHi+bE68M2ko7rf9fGYfw/2gnIBeJ4XlkaPHXaiIQjEuS+qU2ONqhyPZzZ0CD5S7s7dTXJkfudqJkCX/j9g==
X-Complaints-To: Please send complaints to abuse AT motzarella DOT org with full headers
NNTP-Posting-Date: Sun, 20 Jul 2008 00:09:41 +0000 (UTC)
X-Mailer: Mozilla 4.75 [en] (Win98; U)
X-Accept-Language: en
X-Auth-Sender: U2FsdGVkX188h1U7e3rcvwZWYB3644Z5jjg3S8jq+adcyJhs1aKB2A==
Cancel-Lock: sha1:/gDRvkYoc2Do1WN8bdRCusEok50=
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Andris Pavenis wrote:
> 
... snip ...
> 
> The first message of the thread is
>
> http://www.delorie.com/djgpp/mail-archives/browse.cgi?p=djgpp-workers/2007/12/05/17:54:46
> 
> I included there the source of my initial test program used for
> profiling. Remove shuffling addresses and freeing memory, link it
> with nmalloc and You will get what I use for testing when large
> enough number of allocations were specified (like more that
> 1000000, You should have plenty of memory for that). This test
> program randomly crashed when linked with nmalloc and never for
> me if with DJGPP v2.04 own malloc (for this test slowness of
> free() does not matter of course).
> 
> I do not know whether it is possible to reproduce the problem on
> system different from WinXP.

I have heard nothing back since my earlier replies about this
message.  I reported that the system doesn't compile or link,
depending on the parameters given to gcc.

I believe the basic problem is mixing the use of new/dispose with
malloc/free.  This can't be properly evaluated without pulling down
the whole C++ library system, or installing the revised malloc as a
portion of the C++ library.  That won't work if the C++ system
makes malloc depend on new.

As I see it you are generating a herd of mallocs with randomized
sizes, mixing them up, and then freeing all of them.  I have
performed other tests that effectively do this, with no problems. 
If I am correct I am even willing to write a C test routine to do
it, avoiding the mixing problems.

-- 
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: <http://cbfalconer.home.att.net>
            Try the download section.