Re: [PATCH 3/3] Add DWARF index cache

[Simon Marchi <simon.marchi@polymtl.ca>]

On 2018-07-25 14:47, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simon.marchi@ericsson.com> writes:
> 
> Simon> - The cache is disabled by default.  I think once it's been out 
> there
> Simon>   and tested for a while, it could be turned on by default, so 
> that
> Simon>   everybody can enjoy it.
> 
> I think the patch to enable it should also come with a patch to disable
> it in the test suite, since otherwise "make check" is going to fill up
> the cache and maybe have other unintended consequences.

Indeed, the default behavior should probably be to test without the 
index cache.  I also made a board file to test using the index cache.  
It overrides the gdb_load proc to load the binary file twice in GDB (the 
first time to populate the cache, the second time to read from it).  I 
don't think it's very useful, so I didn't include in this series.

> Simon> +static void
> Simon> +set_index_cache_command (const char *arg, int from_tty)
> Simon> +{
> Simon> +  printf_unfiltered ("\
> Simon> +Missing arguments.  See \"help set index-cache\" for help.\n");
> 
> This probably needs _().  I think there were a couple other
> (non-debug-print) cases of this as well.

Right, will fix.

> Simon> +  printf_unfiltered ("The index cache is currently %s.\n",
> Simon> +		   global_index_cache.enabled () ? "enabled" : "disabled");
> 
> Like here.
> 
> Simon> +static void
> Simon> +test_mkdir_recursive ()
> Simon> +{
> Simon> +  char base[] = "/tmp/gdb-selftests-XXXXXX";
> Simon> +
> Simon> +  if (mkdtemp (base) == NULL)
> Simon> +    perror_with_name ("mkdtemp");
> Simon> +
> Simon> +  std::string dir = string_printf ("%s/a/b", base);
> Simon> +  SELF_CHECK (create_dir_and_check (dir.c_str ()));
> Simon> +
> Simon> +  dir = string_printf ("%s/a/b/c//d/e/", base);
> Simon> +  SELF_CHECK (create_dir_and_check (dir.c_str ()));
> 
> It would be nice to clean up afterward.

I'll try to do that.

> Simon> diff --git a/gdb/dwarf-index-write.h b/gdb/dwarf-index-write.h
> [...]
> Simon> +void write_psymtabs_to_index (struct dwarf2_per_objfile
> *dwarf2_per_objfile,
> Simon> +			      const char *dir, const char *basename,
> Simon> +			      dw_index_kind index_kind);
> 
> I tend to prefer the explicit "extern" but I don't know if there's an
> official style.

Actually, the official style is to use extern (I still don't know why).  
I can add it.

> This patch seems fine to me and I like this feature -- I would use it
> for sure.
> 
> I think it needs a doc + NEWS patch.

Oh you're right.  I forgot it when "upgrading" the patchset from RFC to 
RFA.

> I think it would be good to file bugs for the future features you
> mentioned in the email.

Ok.

Thanks for the review, I'll work on a v2.

Simon