jingham added a comment.

In D74136#1871751 <https://reviews.llvm.org/D74136#1871751>, @labath wrote:

> In D74136#1870029 <https://reviews.llvm.org/D74136#1870029>, @jingham wrote:
>
> > In D74136#1869622 <https://reviews.llvm.org/D74136#1869622>, @kwk wrote:
> >
> > > @labath @jingham to summarize from what I read here and what I chatted 
> > > about with @labath , the following  is a possible way to go for now, 
> > > right?
> > >
> > > 1. We're not going to introduce my flag.
> > > 2. You're both not perfectly happy with the way things are documented at 
> > > the moment and dislike some of the implementation as in in LLDB but 
> > > chaning it should not be part of this patch.
> > > 3. @jingham wouldn't want to introduce `--compile-unit` as a flag that 
> > > @labath proposed.
> > > 4. You want `breakpoint set --file` to search everything, that is to say 
> > > compilation units and files referenced in `DW_AT_decl_file`.
> > >
> > >   If you can confirm that this is correct, then I can refactor this patch 
> > > to remove the switch and change the default behavior for `breakpoint set 
> > > --file`. Especially point 4. is important I guess.
> >
> >
> > There's a point (5) which is that the search in 4 should be conditioned on 
> > the setting of the "target.inline-breakpoint-strategy".  That way if people 
> > have big projects but don't ever #include source files, and don't build 
> > with LTO, they can turn off these extra searches, which might end up being 
> > costly and to no purpose for them.
>
>
> I think we're on the same page, but I'll just try to rephrase this in my own 
> words.
>
> Basically, I'm hoping that we could get the `--file+--function` combo to 
> behave the same way `--file+--line`. I.e., that the --file argument should 
> specify the file that the function is defined in (i.e. the DW_AT_decl_file 
> attribute of the function), and not the compile unit (DW_AT_name of the 
> compile unit containing the function). The way the 
> `inline-breakpoint-strategy` setting comes into play is that if the setting 
> value is "headers" and the user specifies a `.cpp` file, then we will assume 
> that the .cpp file also matches the name of one of the compile units, and we 
> will not attempt to search compile units with different names (which is the 
> same thing we do for line breakpoints, I believe).


That is my understanding as well.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74136/new/

https://reviews.llvm.org/D74136



_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits

Reply via email to