X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
X-Recipient: djgpp AT delorie DOT com
X-Recipient: dj AT delorie DOT com
Message-ID: <55434021.4010606@iki.fi>
Date: Fri, 01 May 2015 11:58:09 +0300
From: Andris Pavenis <andris DOT pavenis AT iki DOT fi>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.6.0
MIME-Version: 1.0
To: DJ Delorie <dj AT delorie DOT com>, djgpp AT delorie DOT com
Subject: Re: GCC 5.1.0 problem with <dos.h>
References: <mhmgck$ims$1 AT usenet DOT news DOT interia DOT pl> <201504280005 DOT t3S05t2U020439 AT delorie DOT com> <mhnhb8$m7m$1 AT usenet DOT news DOT interia DOT pl> <201504281718 DOT t3SHIpWZ009266 AT envy DOT delorie DOT com>
In-Reply-To: <201504281718.t3SHIpWZ009266@envy.delorie.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Reply-To: djgpp AT delorie DOT com

On 04/28/2015 08:18 PM, DJ Delorie wrote:
> Ah, I see.
>
> I see no reason to not spin a 2.05 release just to get this out, but
> the only other issue we've had stopping this is the malloc one.  IIRC
> the last concern I had with changing malloc was breaking the ABI, but
> if Juan and Andris can agree on what malloc "works" we'll just go with
> that and have something that can replace both 2.03 and 2.04.
>
>
I have updated nmalloc (sorry for 2 missing headers accidentally left out of first commit),

I also restored src/rmake.cc (missing since v2.02) but used by testsuite. Could it really be true 
that nobody have dries to build testsuite for so long time.

Also some other changes like adding make variables EXTRA_CLAGS and EXTRA_CXXFLAGS so that extra 
flags are visible in make output. Used this to workaround cmalloc() problem (GCC-5.1 compiles it to 
infinite tail recursion at optimization levels 2+).

So library now seems to build OK. Should do some more testing with real software.

I guess we'll need version v2.05 in beta for some time before releasing it as current

Problems:

math,.h:

sincos() has different prototype from GCC builtin and also procedure with same name in glibc. It is 
however GNU extension. Both changing API and leaving it as we have have different disadvantages. I 
do not think sincos() is very widely used, so I would vote for changing prototype

GCC headers (lib/gcc/5.10/include):

does not work nicely with -Wqual-cast. I'll m also getting other warnings like uninitialized 
variable from SSE functions but only when building DJGPP test-suite.

testsuite: there are tests for memory allocation debugging which of course no more buiolds after 
change to nmalloc.



Otherwise: DJGPP v2.04 CVS version build process works nicely with gcc-5.1.0 (same build that is 
uploaded) except for problems with testsuite.

Andris