Re: pgsql: Clarify coding of .exe patch - Mailing list pgsql-committers

From Tom Lane
Subject Re: pgsql: Clarify coding of .exe patch
Date
Msg-id 5139.1099288069@sss.pgh.pa.us
Whole thread Raw
In response to Re: pgsql: Clarify coding of .exe patch  (Neil Conway <neilc@samurai.com>)
Responses Re: pgsql: Clarify coding of .exe patch
List pgsql-committers
Neil Conway <neilc@samurai.com> writes:
> Tom, can you elaborate on why you prefer sizeof()? ISTM that a decent
> compiler will evaluate the strlen() at compile-time if the argument is a
> compile-time constant. GCC will do this (even at -O0!), for example.

sizeof() is defined as a compile-time constant by the C language
specification.  strlen() is not a compile-time constant, and in my
judgement a compiler that evaluates it as such is exceeding its
authority.  The *linker* could in many cases reduce the call to a
constant with certainty, because it could know whether the function
reference was being resolved to the normal C-library definition or not.
(But it would not have very much scope to do anything with the knowledge
:-(.)  gcc's behavior is completely indefensible; it is quite analogous
to my writing some hexadecimal bit-pattern as a substitute for
upper("foo") because I think I know what upper() will behave as.

This is not an argument against our replacing strlen(".exe") by 4
when we know that we intend to invoke a version of strlen() that will
act that way.  I am just pointing out that the C compiler has no
business doing this for us; we'd be squawking very loudly if it made
entirely-similar assumptions that we chanced not to like as much.

            regards, tom lane

pgsql-committers by date:

Previous
From: Neil Conway
Date:
Subject: Re: pgsql: Clarify coding of .exe patch
Next
From: Neil Conway
Date:
Subject: Re: pgsql: Clarify coding of .exe patch