This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [RFA] Fix mi-break.exp: 'b "basics.c":16'
On Thu, May 09, 2002 at 07:31:17PM -0400, Elena Zannoni wrote:
> Daniel Jacobowitz writes:
> > On Thu, May 09, 2002 at 03:40:42PM -0700, Michael Snyder wrote:
> > > Daniel Jacobowitz wrote:
> > > >
> > > > [I dislike decode_line_1. But that seems to be the general consensus...]
> > > >
> > > > There was a logic error in the code to handle "file":line. Several,
> > > > actually. I suspect that it once worked and had bitrotten as the behavior
> > > > of the function changed. For instance, at the beginning of the function:
> > > >
> > > > if (p[0] == '"')
> > > > {
> > > > is_quote_enclosed = 1;
> > > > (*argptr)++;
> > > > p++;
> > > > }
> > > >
> > > > Then below it checked 'is_quote_enclosed && (**argptr) == '"').
> > >
> > > [line break inserted]
> > >
> > > > That'll only be true given a literal '"":<line>', which was not
> > > > the intent of the test.
> > >
> > > I don't understand this statement. Between the two bits of text
> > > you've named, both p and *argptr are changed. Seems to me, there
> > > could be any number of characters between the two quotes.
> >
> > This is the part that took me the longest to work out in the first
> > place, I think. `p' being changed doesn't matter here; it should end
> > up point at (before?) the colon.
> >
> > The first if statement is at line 630. The second is at 929. Between
> > the two, *argptr is only changed if we found a C++ class. It might be
> > changed on "interestingly" named files, also, but if so it's purely a
> > bug in the twistiness of linespec. At line 633 it points to the
> > _second_ character (right after the first quote), so if **argptr ==
> > '"', that means the second quote is right after the first.
>
> As usual, I had to go and see when/why the test started failing,
> and it was because a patch from March 2001, which was fixing
> a core dump on the following:
>
> (gdb) break "foo"
>
> I have verified that your patch doesn't reintroduce the core dump.
> I guess we can check that in.
OK, committed.
> Hey, any chance you can add a break "foo" in some testfile somewhere?
> Probably break.exp?
Just posted one.
>
> Elena
>
>
> >
> > >
> > >
> > > > This patch updates the behavior, should change nothing else, and causes no
> > > > regressions. OK to check in?
> > > >
> > > > --
> > > > Daniel Jacobowitz Carnegie Mellon University
> > > > MontaVista Software Debian GNU/Linux Developer
> > > >
> > > > 2002-04-02 Daniel Jacobowitz <drow@mvista.com>
> > > >
> > > > * linespec.c (decode_line_1): Check for a double quote after
> > > > a filename correctly.
--
Daniel Jacobowitz Carnegie Mellon University
MontaVista Software Debian GNU/Linux Developer