Re: incorrect error locations for command-line code

[Aharon Robbins]

Andy,

Hi. I've applied the second patch to my sources and added a test.

Work is in progress on getting a public read-only CVS archive set up. No
news yet, but I'll announce it when it's available.

Thanks!

Arnold

> Date: Tue, 05 Sep 2006 14:57:44 -0400
> From: "Andrew J. Schorr" <address@hidden>
> Subject: incorrect error locations for command-line code
> To: address@hidden
>
> --sdtB3X0nJg68CQEu
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: inline
>
>
> Hi,
>
> It seems that using a mix of code supplied with '-f' and --source
> can lead to errors being reported with the wrong location (using
> gawk version 3.1.5).  The problem is best understood with some examples:
>
> This is correct:
>
>    bash-3.00$ ./gawk 'BEGIN {return junk}'
>    gawk: BEGIN {return junk}
>    gawk:        ^ `return' used outside function context
>
> This is also correct:
>
>    bash-3.00$ ./gawk --source 'BEGIN {return junk}'
>    gawk: BEGIN {return junk}
>    gawk:        ^ `return' used outside function context
>
> But this is wrong:
>
>    bash-3.00$ ./gawk -f /dev/null --source 'BEGIN {return junk}'
>    gawk: /dev/null:1: BEGIN {return junk}
>    gawk: /dev/null:1:        ^ `return' used outside function context
>
> As you can see, the error is incorrectly reported as occurring inside
> /dev/null.  It seems that awkgram.y:get_src_buf fails to reset the
> source & sourceline global variables correctly when it encounters
> the CMDLINE code.
>
> A possible patch is attached.  Note that this patch changes current
> behavior slightly, in that it causes the original (correct) behavior
> above to change slightly to include a "cmd. line:1" label.  Here are the
> examples with this patch in place:
>
>    bash-3.00$ ./gawk 'BEGIN {return junk}'
>    gawk: cmd. line:1: BEGIN {return junk}
>    gawk: cmd. line:1:        ^ `return' used outside function context
>
>    bash-3.00$ ./gawk --source 'BEGIN {return junk}'
>    gawk: cmd. line:1: BEGIN {return junk}
>    gawk: cmd. line:1:        ^ `return' used outside function context
>
>    bash-3.00$ ./gawk -f /dev/null --source 'BEGIN {return junk}'
>    gawk: cmd. line:1: BEGIN {return junk}
>    gawk: cmd. line:1:        ^ `return' used outside function context
>
> If that behavior is not desirable (to include the "cmd. line:1" location
> even when the command-line code does not included an embedded newline),
> then the 2nd patch can be used.
>
> Regards,
> Andy
>
>
> #####################################################################################
> This Mail Was Scanned by 012.net AntiVirus Service2- Powered by TrendMicro 
> Interscan
> --sdtB3X0nJg68CQEu
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: attachment; filename="errloc1.patch"
>
> --- awkgram.y.errloc  2005-07-26 14:07:43.000000000 -0400
> +++ awkgram.y 2006-09-05 14:48:41.162413000 -0400
> @@ -1271,7 +1271,6 @@ again:
>               }
>               if (srcfiles[nextfile].val[l-1] == '\n') {
>                       /* has terminating newline, can use it directly */
> -                     sourceline = 1;
>                       lexptr = lexptr_begin = srcfiles[nextfile].val;
>                       /* fall through to pointer adjustment and return, below 
> */
>               } else {
> @@ -1295,6 +1294,8 @@ again:
>                       /* set vars and return */
>                       lexptr = lexptr_begin = buf;
>               }
> +             source = NULL;
> +             sourceline = 1;
>               lexend = lexptr + l;
>               nextfile++;     /* for next entry to this routine */
>               return lexptr;
>
> --sdtB3X0nJg68CQEu
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: attachment; filename="errloc2.patch"
>
> --- awkgram.y.errloc  2005-07-26 14:07:43.000000000 -0400
> +++ awkgram.y 2006-09-05 14:55:04.290127000 -0400
> @@ -1272,6 +1272,7 @@ again:
>               if (srcfiles[nextfile].val[l-1] == '\n') {
>                       /* has terminating newline, can use it directly */
>                       sourceline = 1;
> +                     source = NULL;
>                       lexptr = lexptr_begin = srcfiles[nextfile].val;
>                       /* fall through to pointer adjustment and return, below 
> */
>               } else {
> @@ -1293,6 +1294,8 @@ again:
>                       buf[++l] = '\0';
>  
>                       /* set vars and return */
> +                     sourceline = 0;
> +                     source = NULL;
>                       lexptr = lexptr_begin = buf;
>               }
>               lexend = lexptr + l;
>
> --sdtB3X0nJg68CQEu--
>