Messages in this thread | | | Date | Thu, 8 Mar 2012 08:48:37 +0100 | From | Ingo Molnar <> | Subject | Re: [PATCH] perf: Incorrect use of snprintf results in SEGV |
| |
* Peter Seebach <peter.seebach@windriver.com> wrote:
> On Wed, 7 Mar 2012 21:37:25 +0100 > Ingo Molnar <mingo@elte.hu> wrote: > > > You are missing two important aspects: > > > > 1) Dynamic reallocation on snprintf() failure is an utterly rare > > thing - it is used in less than 1% of snprintf() invocations. > > (Yes, I just checked a couple of codebases.) > > I would agree that it's very rare. But then, using the return > value at all isn't especially common in my experience -- the > only interesting part, most of the time, is "we're sure this > didn't overrun the buffer".
Erm. Doing:
+= snprintf(...);
is a *very* common pattern within the kernel. It occurs more than a thousand times - i.e. about 25% of all snprintf uses (~5000 instances) within the kernel does care about the return value.
I found only a single case that did a reallocation if the buffer did not fit. Lets assume that I missed some and there's 4 altogether.
I.e. the API usage proportion, within the kernel project, looks like this, approximately:
snprintf() call site that:
does not care about the return value: 75.0% uses the return value as a 'written' count: 24.9% wants to dynamically reallocate: 0.1%
> > We *DONT* want to make APIs more fragile just to accomodate a > > rare, esoteric usecase! > > I would view snprintf as an API which already exists.
Changing it is obviously not possible anymore.
I was just countering your justification for it - which is still wrong. People might read that and use it to justify newly introduced, crappy APIs.
The 0.1% usecase is absolutely not a valid excuse to make an API less robust - *especially* when a separate API could serve that 0.1% case just fine.
When designing APIs it is of utmost importance how average developers intuitively *think* it works - not how the designer thinks it should work ... Any severe mismatch between the two is a serious design FAIL that should not be repeated in new code.
Thanks,
Ingo
| |