strtoul pitfalls

strtoul pitfalls

[Bruno Haible]

Hi Alejandro,

reviewing the strtoul() calls in gnulib and gettext, I found several
places where the original coder had fallen into one of these two pitfalls.
I would find it good to mention both pitfalls more prominently in the
man-pages (the first one is mentioned as a NOTE, should better be called
a BUG).

* Leading whitespace and a sign are accepted by default.
  Often this is not desired.
  To avoid this, the caller should pre-parse the initial portion
  of the string, up to the first digit.

* If values > MAX should lead to a diagnostic, they arise in two
  circumstances:
    - errno = ERANGE,
    - errno = 0, end == str, but value > MAX.
  It is not obvious how to arrange the code, so as to get the same
  diagnostic, without code duplication.

Bruno

[1] https://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=008ff109def9bdfdd04a12355d8f4a835045ae41
[2] https://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=3df6da258c338e4fd165a2850bb8d4b0f39b6e9b
[3] https://git.savannah.gnu.org/gitweb/?p=gettext.git;a=commitdiff;h=b8ab4f2bef0ff18cf1ae749c1c81a359a40d0450

Re: strtoul pitfalls

[Alejandro Colomar]

[-- Attachment #1: Type: text/plain, Size: 1617 bytes --]

Hi Bruno,

On Fri, Mar 21, 2025 at 08:05:18PM +0100, Bruno Haible wrote:
> Hi Alejandro,
> 
> reviewing the strtoul() calls in gnulib and gettext, I found several
> places where the original coder had fallen into one of these two pitfalls.
> I would find it good to mention both pitfalls more prominently in the
> man-pages (the first one is mentioned as a NOTE, should better be called
> a BUG).
> 
> * Leading whitespace and a sign are accepted by default.
>   Often this is not desired.
>   To avoid this, the caller should pre-parse the initial portion
>   of the string, up to the first digit.
> 
> * If values > MAX should lead to a diagnostic, they arise in two
>   circumstances:
>     - errno = ERANGE,
>     - errno = 0, end == str, but value > MAX.
>   It is not obvious how to arrange the code, so as to get the same
>   diagnostic, without code duplication.

Agree.  I think this page needs a long BUGS and CAVEATS pair of
sections.  I'll write something about it.  Feel free to report more bugs
about these APIs; this will help kill them soon.  I'll present the
resulting manual pages as part of the defense of my paper for C2y.


Have a lovely night!
Alex

> 
> Bruno
> 
> [1] https://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=008ff109def9bdfdd04a12355d8f4a835045ae41
> [2] https://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=3df6da258c338e4fd165a2850bb8d4b0f39b6e9b
> [3] https://git.savannah.gnu.org/gitweb/?p=gettext.git;a=commitdiff;h=b8ab4f2bef0ff18cf1ae749c1c81a359a40d0450
> 
> 
> 

-- 
<https://www.alejandro-colomar.es/>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]