[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #PUM-976132]: Compile problems with NetCDF 3.6.3 and gcc's

Hi Sean,

Thanks for looking at the problem and providing your comments.

I will use strlcat() instead, which I will provide a version of
if it's not available on the target platform.

> This is buggy too. The first time through the while loop 'sp' points to
> 'newlen' bytes of memory. Assume 'repls[j]' is a very long string. The
> strncat() will fill 'sp' and leave it unterminated. So 'newlen-1' would
> be better. But that's still no good because you advance the 'sp'
> pointer, so each time through the loop you should be passing a smaller
> and smaller length to strncat(). I suppose there is no real danger
> since you compute newlen correctly in the first place, but still, if
> you're not passing the correct length to strncat(), you may as well just
> use strcat. :)

I have to disagree with this analysis, but I didn't provide the
context in which the code fragment was embedded, so I can understand
why you concluded their may be a bug.

First, repls[j] is always a short string of known length lens[j], as
these two arrays are precomputed on initialization in a table of
replacement strings and their lengths (for some special characters that
might appear in names).  Thus strncat() cannot fill sp and leave it
unterminated, as the value of newlen is explicitly computed to include
the lengths of all replacement strings + 1 for the null termination.

Also sp is advanced in the loop by exactly the same value that newlen
is decremented by

        sp += lens[j];
        newlen -= lens[j];

so a smaller and smaller value of newlen is passed to strncat() for
each iteration.

Of course that's irrelevant to this strncat() call, because it
aborted on the first time through the loop, when sp pointed to a
null byte at the beginning of an allocated buffer for a string.
Also changing the strncat(...,newlen) to strncat(...,newlen-1)
led to the same abort on the first time through the loop.

But I appreciate the suggestion to use strlcat() and will make that
change this afternoon.  Unfortunately, -D_FORTIFY_SOURCE=2 doesn't
seem to check strlcat() calls, so I'd better get it right!

Thanks again,

--Russ


Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: PUM-976132
Department: Support netCDF
Priority: Normal
Status: Closed