NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/sys/hp/10048

Internet Message Format | 1992-09-04 | 2.4 KB

Path: sparky!uunet!dtix!darwin.sura.net!Sirius.dfn.de!chx400!sicsun!slhp1.epfl.ch!lecom
From: [e-mail] (Claude Lecommandeur)
Newsgroups: comp.sys.hp
Subject: Re: that HPUX cc compiler
Message-ID: <[e-mail]>
Date: 4 Sep 92 07:34:11 GMT
References: <[e-mail]> <[e-mail]>
Sender: [e-mail]
Organization: Ecole Polytechnique Federale de Lausanne
Lines: 40

In article <[e-mail]>, [e-mail] (Paul Hite) writes:
|>
|> HP gets a lot of grief from the behavior of their c optimizer.  I'm sure
|> that some of it is valid.  But a lot of the posts basicly say something
|> like:
|>     I've got a program that works without the optimizer and fails
|>     when I optimize it.  Therefore HP's optimizer is broken.
|>
|> Here's a real live counter-example to that argument.  I wrote a replacement
|> to w and it worked without the optimizer and failed when I optimized it.
|> I reduced the program to a tiny example that also failed with the optimizer.
|> The complete program is below my signature, but here is the important part:
|>     struct utmp entry;
|>     while(fread(entry, sizeof entry, 1, u)) {
|>         bogus = entry.ut_time;
|> With the optimizer, bogus always had the same value.  Without the optimizer
|> various correct values are printed.  The problem is the fread.  It should be:
|>     while(fread(&entry, sizeof entry, 1, u)) {
|>                 ^
|> And the optimizer works fine on the corrected program.  Actually it's amazing
|> that the program works at all the first way.  All of this is on an 847, those
|> of you on other platforms may get different results with the buggy program.
|> (Indeed, Steinar Haug found the bug for me as he struggled to get my program
|> working on a 425!)

    I have seen this behaviour before, i mistakenly gave the actual structure
instead of its address as a parameter to a function call and it worked when
compiled without -O and broke with -O. (it's hard to debug because it works
with -g). In this case, it is definetely not the optimiser which is broken,
but the compiler itself. It looks like it tries to be too much clever.

--


                    Claude Lecommandeur
                    Service Informatique Central
                    Ecole Polytechnique Federale de Lausanne
                    1015 LAUSANNE (SWITZERLAND)
                    E-Mail : [e-mail]
                    Tel : [phone]