Message-Id: <199804281259.OAA30744@ieva06.lanet.lv>
From: "Andris Pavenis" <pavenis AT lanet DOT lv>
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Date: Tue, 28 Apr 1998 14:58:11 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: Patch to add target specific cleanup for gcc.c
CC: djgpp-workers AT delorie DOT com
References: <199804271144 DOT NAA33208 AT ieva06 DOT lanet DOT lv>
In-reply-to: <Pine.SUN.3.91.980427160932.23790A-100000@is>
Precedence: bulk

Date sent:      	Mon, 27 Apr 1998 16:14:50 +0300 (IDT)
From:           	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Subject:        	Re: Patch to add target specific cleanup for gcc.c

> 
> On Mon, 27 Apr 1998, Andris Pavenis wrote:
> 
> > I think that creating one extra file can still be in average faster than opening EVERY 
> > temporary files once more. Of course this is safer than approach I used,
> > but is it worth the overhead we'll have
> 
> We clearly disagree on that, but since the port of GCC 2.8.1 is your 
> work, you will have to decide.
> 
> Here's my opinion: I think it is not clean for GCC to work around bugs in
> the library, real or potential, unless those library bugs cannot be fixed
> easily.  In this case, IMHO the fix is easy and should be applied to the
> library.

Perhaps I can use modified version of mkstemp(). I tried to test how much
it slows down things (of course not with gcc) and the slowdown was not very
significant when I run strip with output to temporary file generated by mkstemp()
and closed imediatelly in comparisson with names generated by mktemp()

Anyway it's better to modify these procedures specially for gcc but in this way
these procedures will be no more almost totally broken as in previos approch.
It is not recommended to modify gcc common part unless it is really necessary.
So I think there will be no danger in getting port of the future versions of gcc.

I think we could also provide possibility to turn off this feature (ON by default)
by some environment parameter for example:
	SET  GCC_USE_MKSTEMP=n
turns off this feature (perhaps the default should be y) 

Andris