Re: Patch: fix various issues found by static analysis [infokey]

[Gavin Smith]

On Tue, Oct 15, 2024 at 12:36:59PM +0200, Vitezslav Crhonek wrote:
> From aae25e4335fa4127b9c3e73486095df304ae735a Mon Sep 17 00:00:00 2001
> From: Vitezslav Crhonek <vcrhonek@redhat.com>
> Date: Tue, 15 Oct 2024 10:58:52 +0200
> Subject: [PATCH 2/7] * info/infokey.c: add initializer
> 
> ---
>  info/infokey.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/info/infokey.c b/info/infokey.c
> index ec06cec51f..a42f896d3f 100644
> --- a/info/infokey.c
> +++ b/info/infokey.c
> @@ -208,7 +208,7 @@ compile (FILE *fp, const char *filename, int 
> *suppress_info, int *suppress_ea)
>    int oval = 0;
>    char comment[10];
>    unsigned int clen = 0;
> -  int seq[20];
> +  int seq[20] = { 0 };
>    unsigned int slen = 0;
>    char act[80];
>    unsigned int alen = 0;

I'm guessing this does not fix a bug, but the fact that the function is
a "state machine" implemented with a switch statement inside a loop
makes it difficult to immediately confirm that 'seq' is not read before
it is written to, in code such as

  /* Only allow "1 menu-digit".  (This is useful if
     this default binding is disabled with "#stop".)
     E.g. do not allow "b menu-digit".  */
  if (seq[0] != '1' || seq[1] != '\0'
      || section != info)
    {
      syntax_error (filename, lnum,
             _("cannot bind key sequence to menu-digit"));  
    }

When checking over the code I found a bug that if there were exactly
20 characters in 'seq', then the array would not be null-terminated
for the call to keymap_bind_keyseq, which relies on it being null-terminated.

Hence I fixed this by initialising seq to have an initial 0, and also
always null-terminating the sequence.  This may also stop your static
analysis warning.