Re: [PATCH v3 3/3] gdb: extend the [[N]]::foo syntax for files

[Kevin Buettner <kevinb@redhat.com>]

On Thu, 10 Jul 2025 12:07:48 -0300
Guinevere Larsen <guinevere@redhat.com> wrote:

> On 7/9/25 11:27 PM, Kevin Buettner wrote:
> > On Thu, 19 Jun 2025 15:05:40 -0300
> > Guinevere Larsen <guinevere@redhat.com> wrote:
> >  
> >> +    gdb_test "print ${ns0}::'${::srcfile_lib}'::gdb_dlmopen_glob" \
> >> +	"No function .${::srcfile_lib}. in specified context."  
> > I'm wondering about this test.  ${::srcfile_lib} refers to a file,
> > right? If so, why does the message say "function"?  
> 
> This is an annoying case to work with... The error "No function %s in 
> specified context" comes from the "block: block COLONCOLON name" rule, 
> around line 1105 in my tree.
> 
> The issue here is that the function "classify_name" couldn't find a file 
> named ${::srcfile_lib} in the specified namespace, and so it classified 
> that string as a generic name, rather than a file name. It's not even a 
> factor of not having enough information to differentiate, the issue is 
> that the name is miscategorized it would be impossible for us to even 
> know when it was meant as a file or a function.
> 
> At present, this error can only refer to a file if the user is searching 
> for something in a namespace, so I could make that error checking 
> slightly more thorough, and keep the current message when no namespace 
> ID was used, and have a new "No function or file %s in specified 
> context" message. Would that be better, in your opinion?

While it would be nice to use "file" to refer to files and "function"
for functions, if we really can't differentiate, then I think using
"No function or file %s..." is better than the current situation.

Kevin