Re: D7 Compiler Bug

From: Marco van de Voort (marcov_at_stack.nl)
Date: 10/07/04 

Date: Thu, 7 Oct 2004 12:17:38 +0000 (UTC)

On 2004-10-07, VBDis <vbdis@aol.com> wrote:
>>> data type. The Delphi compiler IMO is very reliable, so it's very strange
>>> to learn that there exist situations where the compiler definitely
>>> produces broken code :-(
>>
>>IMHO this is a documentation fault, and the compiler is fine. FPC
>>independantly has the same behaviour.
>
> A proper implementation only would require an "legacy" switch, or another
> mode in FPC, that enables Object and disables AnsiString and other
> unsupported extensions in that context.

IMHO this is a kludge. If you want static objects that badly, define a new
object system for it, don't try to retrofit legacy systems.

> Then the compiler can reject unhandled situations, instead of silently
> producing buggy code.

There is no official code that does this. Moreover, IMHO the code was doubly
broken, since an object was accessed before creation.

>>It's a pity that it hasn't been released to PD (or MPL) yet, but I still
>>want to try to get Turbo Vision compiling under Delphi one day. That would
>>really stress the object mechanism.
>
> During my first steps with BC3 I studied the OWL and found it not very well
> designed, with regards to customization. GUI elements deserve a very careful
> design, that does not hard code a specific behaviour of sub-components into the
> base classes. In my experiments it was impossible to preserve the current
> position in lists when scrolling up and down by pages - the current position
> frequently changed by arbitrary numbers of lines, depending on the situation.
> Perhaps newer versions have a better design?

I never used OWL or any other Borland Windows compiler before D4. Only TV,
LCL and a bit of VCL.

TV was reasonably well designed IIRC, except for some 16-bitisms that are
easily fixed. Most problems stem from limitations in the TP object model,
not from TV.

Relevant Pages

  • Re: How long would it take to build Win64/Unicode Delphi?
    ... Making a commercial fork essentially. ... I'am not sure if all people would have contributed if FPC were BSD ... tight IDE integration requires that Borland ports their compiler. ... The FPC design is mainly concerned about maintainability and portabilitiy. ...
    (borland.public.delphi.non-technical)
  • Re: why still use C?
    ... OOP in general tends to be slower. ... As if the compiler didn't know... ... > their associated method functions from a design perspective. ... as a language design I feel it has failed. ...
    (comp.lang.c)
  • Re: Will Delphi will do this one day?
    ... > to think about compiler optimization. ... these guys have done what Borland should have done years ... are always arrranged for profit maximization, ... > functionality than FPC offers. ...
    (borland.public.delphi.non-technical)
  • Re: writeObject signature
    ... > or the compiler must do a cast to Object. ... Java refers to a syntax for type conversion -- namely, ... it would be wrong (or bad design) to ...
    (comp.lang.java.programmer)
  • Re: Inside an FBI Computer Forensics Lab
    ... You should be comparing a hardware engineer who puts ... design done in a hardware description language. ... My compiler is open source also. ... The attack involves compromising the compiler before the hash ...
    (alt.privacy)