Re: [RFA] Allow setting breakpoints on inline functions (PR 10738)

[Gary Benson <gbenson@redhat.com>]

Tom Tromey wrote:
> >>>>> "Gary" == Gary Benson <gbenson@redhat.com> writes:
> 
> Gary> This patch, which applies on top of Tom's ambiguous linespec work,
> Gary> allows you to set breakpoints on inlined functions.  Although it
> Gary> can't be committed until Tom's stuff goes in, I'm posting it for
> Gary> feedback now.
> 
> I noticed that the manual node "Inline Functions" says:
> 
>        There are some ways that GDB does not pretend that inlined function
>     calls are the same as normal calls:
> 
>        * You cannot set breakpoints on inlined functions.  GDB
>          either reports that there is no symbol with that name, or
>          else sets the breakpoint only on non-inlined copies of the
>          function.  This limitation will be removed in a future
>          version of GDB; until then, set a breakpoint by line number
>          on the first line of the inlined function instead.
>     [...]
> 
> I think this needs a small update.

I removed that entire bullet point.

> Also I think this feature deserves a NEWS entry.

I added "* GDB can now set breakpoints on inlined functions." on the
branch.

> I checked it out and played with it a little.  I found one little
> bug.  Using the inline-break test case from the patch:
> 
> (gdb) p &func1
> $1 = (int (*)(int)) 0x4003d8 <main+8>
> 
> That is, it chooses the location of the inline function as the
> address of the function when evaluating an expression.  I think
> this is wrong.  Instead, it should ignore inline instances here,
> returning the address of the out-of-line instance.  And, if there
> is no out-of-line (as in this case), it should error.
> 
> I think one possible way to do this would be to put a flag on
> symbols, marking inline instances, and then have ordinary symbol
> lookup ignore such symbols.  I am not sure how hard this would be.
> There might also be other approaches.

There already is such a flag: SYMBOL_INLINED (sym).  I've pushed
a fix to the branch to make lookup_block_symbol skip over inlined
symbols.  It works as you ask with no regressions.

I won't make a new patch yet as I need to look at some of Jan's
suggestions.

Thanks,
Gary

-- 
http://gbenson.net/