On Mon, Jan 11, 2016 at 9:56 AM, Paul Eggert wrote:
> Andrei Borzenkov wrote:
>>
>> It looks like assignment to first_long_opt is
>> missing somewhere.
>
>
> It seems to have been missing ever since it was added.
>
> I assume that both bugs you mentioned don't occur in glibc. Perhaps all we
> need
Andrei Borzenkov wrote:
It looks like assignment to first_long_opt is
missing somewhere.
It seems to have been missing ever since it was added.
I assume that both bugs you mentioned don't occur in glibc. Perhaps all we need
to do is to merge glibc's argp changes into gnulib. Attached is a fir
Hong Xu wrote:
The BLOCKSIZE is defined as 32768, which may be too large if the default
signed integer size is no larger than 16 bits.
GNU code as a rule does not worry about machines where 'int' is less than 32
bits. POSIX requires 'int' to be at least 32 bits nowadays.
The BLOCKSIZE is defined as 32768, which may be too large if the default
signed integer size is no larger than 16 bits.
---
lib/md2.c| 5 +++--
lib/md4.c| 5 +++--
lib/md5.c| 5 +++--
lib/sha1.c | 5 +++--
lib/sha256.c | 5 +++--
lib/sha512.c | 5 +++--
6 files changed, 18 insertions
Found by Coverity scan for GRUB.
hol_entry_help dereferences state parameter without checking for NULL:
if (first_long_opt || uparams.dup_args)
arg (real, "=%s", "[=%s]", state->root_argp->argp_domain,
stream);
It is called in several places where "st
This was found by Coverity scan of GRUB
hol_entry_help contains the following code
/* A real long option. */
{
int first_long_opt = 1;
__argp_fmtstream_set_wmargin (stream, uparams.long_opt_col);
for (opt = real, num = entry->num; num > 0; opt++, num--)
if (opt
