(In reply to comment #0)
> ... which is totally already part of posix.  but in this case it allocates
> fixed-size buffers on the stack, so it'll false-match on strings over 512 bytes
> long.  hooray!

So is the desired fix to rework strcasecmp to not allocated fixed size buffers,
or would we rather remove it and require a strcasecmp be present in libc?

(In reply to comment #1)
> So is the desired fix to rework strcasecmp to not allocated fixed size buffers,
> or would we rather remove it and require a strcasecmp be present in libc?

both:

- use the operating system's strcasecmp where present
- use a non-broken version (probably from BSD libc) where not present

where?

hot comment action

more comments

Created attachment 5003 [details] [review]
Fix

The patch checks for strcasecmp in configure.ac and if it isn't there, defines
NEED_STRCASECMP.

In xkbmisc.c, if this is defined, the casecmp function (_XkbStrCaseCmp) is
replaced with a new version (from the freeBSD implementation).	If strcasecmp
is already present, this is #define'd to strcasecmp.

Is this the right way to do it?  Or would it be better to change all instances
of _XkbStrCaseCmp to strcasecmp and do it that way?

even better, just removed all use of _XkbStrCaseCmp

This patch appears to be bogus.  tolower() works with
single characters, not strings:

xkbtext.c:1428: warning: passing argument 1 of 'tolower' makes integer from
pointer without a cast
xkbtext.c:1428: warning: cast from pointer to integer of different size
xkbtext.c:1428: warning: cast to pointer from integer of different size
xkbtext.c:1428: warning: initialization makes integer from pointer without a cast

that patch wasn't applied, so, this is still closed.

I wasn't referring to the patch attached to this bug.
The problem is in the patch that went to lib/xkbfile
two days ago:

 revision 1.7
 date: 2006-03-25 23:15:58 +0000;  author: daniels;  state: Exp;  lines: +5 -5
 Bug #3819: Change open-coded _XkbStrCaseCmp to strcmp + tolower.

The new code passes a pointer argument to tolower() in
several places:

 1.7          (daniels  25-Mar-06):          else if
(strcmp(tolower(tok),"layout") == 0)
 1.1          (kaleb    14-Nov-03):              headingtype = HEAD_LAYOUT;
 1.7          (daniels  25-Mar-06):          else if
(strcmp(tolower(tok),"variant") == 0)
 1.1          (kaleb    14-Nov-03):              headingtype = HEAD_VARIANT;
 1.7          (daniels  25-Mar-06):          else if
(strcmp(tolower(tok),"option") == 0)


It also breaks building xserver because there are still
references to _XkbStrCaseCmp() in the xorg tree.

unbroken in HEAD

Fixed as of yesterday, thanks.