From:	Kevin Ashley <k DOT ashley AT ulcc DOT ac DOT uk>
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: Infinite loop???
Date:	Mon, 13 Jul 1998 16:29:20 +0100
Organization:	Posted via ULCC Internet Services
Lines:	28
Message-ID:	<35AA27D0.41C6@ulcc.ac.uk>
References:	<35A3D8C3 DOT 208E6D23 AT cyberdude DOT com> <35A5FC72 DOT C235F6D9 AT eik DOT bme DOT hu>
	<VqrTZAA4qlp1Ew16 AT foobar DOT co DOT uk> <35A7A838 DOT 157A99A3 AT alcyone DOT com> <10o7uMA7JLq1EwoL AT foobar DOT co DOT uk>
NNTP-Posting-Host:	silver.ulcc.ac.uk
Mime-Version:	1.0
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp

Paul Shirley wrote:
> 
> However I am unaware of any x86 (or 68K or MIPS) compiler that does not
> perform arithmetic shifts (and I'm aware of enough apps that would fail
> it this changed) so its a reasonable assumption to make.
> 
> Presumably this is undefined to cope with some of the 'challenged' old
> hardware out there incapable of efficient signed shifts.

No, it could just as well be there because there were already two
established conventions for what was the 'correct' behaviour when
the ANSI standard came into force - but I confess I don't know.
But your assumptions strike me as dangerous. I'm burnt often enough
porting code that makes incorrect assumptions about the constancy of
behaviour of C shift operators to never want to rely on them.

This bug is is particularly problematic (to my mind at least) because
it's difficult to spot from a cursory reading of the code whether or
not it depends on the behaviour of the shift operators in undefined
circumstances. Just because you find that the machines you deal
with currently all behave the same way, sooner or later that assumption
will return to bite you.
  
------------------------------------------------------------------------------
Kevin Ashley                                       K DOT Ashley AT Ulcc DOT ac DOT uk
Special Projects Manager             http://www.ulcc.ac.uk/staff/Kevin+Ashley
ULCC                        ...ukc!ncdlab!K.Ashley (but probably not any more)
                      This is not a signature

- Raw text -