This is the mail archive of the mailing list for the libstdc++ project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: verbose terminate() on by default

On Tue, Dec 24, 2002 at 01:52:58AM +0100, Gabriel Dos Reis wrote:
> Phil Edwards <> writes:
> | Replacing it with write() would be a bad step, I think.  Now we have
> | multiple system calls, instead of a single one done after formatting.
> I disagree that this will be a bad steo for the reasons you gave.
>   (1)  How many system calls are there?

3 instead of 1

>   (2)  We're not on the critical path.

I agree.  Daniel pointed out that we're terminating, after all.
My efficiency argument is withdrawn.

> | +#define errwrite(s)  write(2,s,strlen(s))
> Is there any reason why this isn't an inline function?

I've actually replaced this with the fragment in, which handles
the case where unistd.h is not necessary.  I intend to match whatever is
used in, for consistency.  (Changing this /and/ to use
inlines is certainly possible, but for a different patch.)

> If we're going to be verbose, then please let's be explanatory.  Also,
> it was agreed at some point that "`...'" is bad quoting.

I just maintained the quoting that was there.  The `foo' style always
bothered me, so if it's been agreed that `foo' is bad, I'll gladly get
rid of it.

I'll be stepping out for some hours, but will provide a new patch when
I return.


I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
                                                 - Edsger Dijkstra, 1930-2002

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]