Re: Weird behaviour with intent

In article <1hrer10.1rpjzoetyakxkN%nospam@xxxxxxxxxxxxx>,
Richard E Maine <nospam@xxxxxxxxxxxxx> wrote:

This is not an error that
compilers are required to catch. Nor is it an error that I'd expect most
compilers to be able to catch.

This is true, however, we have an option to our compiler which says
"don't assume that arguments don't alias." It has several friends, and
when a customer sees a wrong answer, the first thing we do (or the
customer does, if they've read the docs) is compile with all of these
options, which will quickly lead them to the type of bug.

We added this option after we noticed that the Polyhedron benchmarks
contained one example of illegal aliasing. (It's fixed now.)

-- greg
(employed by, not speaking for, QLogic/PathScale)
.

Relevant Pages

  • Re: An Ada Advice Inquiry
    ... Don't blame the language if you are force to ... about how languages and compilers get chosen, ... Complain about Borland, not about C+ ... convince customer to scrap Borland, even less to switch to Ada. ...
    (comp.lang.ada)
  • Re: why doesnt this compile ?
    ... > compilers I tried. ... maybe I'm not entirely out of grump mode yet.) ... Richard Maine ...
    (comp.lang.fortran)
  • Re: Fortran Error Reporting Requirements
    ... The top 3 features the typical customer desires are: faster execution, ... If there was money in it, those compilers that produce fast ...
    (comp.lang.fortran)
  • Re: behavior-preserving optimization in C, was compiler bugs
    ... Most C compilers will compile these two modules without complaint; ... the program works without optimization because the ... program crash or even misbehave. ... A software crash in the field typically means the customer ...
    (comp.compilers)
  • Re: D2005 and D7
    ... >> existing code, old compilers are needed. ... been delivering code and EXEs ... long time, they sometimes needs maintenance. ... customer may want someone else to do it. ...
    (borland.public.delphi.non-technical)