This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA/DWARF] Set enum type "flag_enum" and "unsigned" flags at type creation.
- From: Mark Wielaard <mjw at redhat dot com>
- To: Joel Brobecker <brobecker at adacore dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Thu, 27 Feb 2014 11:30:31 +0100
- Subject: Re: [RFA/DWARF] Set enum type "flag_enum" and "unsigned" flags at type creation.
- Authentication-results: sourceware.org; auth=none
- References: <1390796357-3739-1-git-send-email-brobecker at adacore dot com> <1392820455 dot 21975 dot 235 dot camel at bordewijk dot wildebeest dot org> <1392823115 dot 21975 dot 238 dot camel at bordewijk dot wildebeest dot org> <20140226183157 dot GF4348 at adacore dot com>
On Wed, 2014-02-26 at 10:31 -0800, Joel Brobecker wrote:
> I finally had some time to test this patch, and unfortunately,
> it does introduce some regression (in Ada). For instance in
> homonym.exp:
Sorry, I didn't have gcc-gnat installed and so missed this. I have it
installed now. BTW. Are there any overviews of what are expected
results? For make check RUNTESTFLAGS='--directory gdb.ada' I get:
=== gdb Summary ===
# of expected passes 490
# of unexpected failures 29
# of unexpected successes 8
# of expected failures 2
# of known failures 1
# of unsupported tests 3
Is that reasonable? The amount of failures seems a bit high. The
testsuite is not supposed to be (near) zero-fail?
> type Integer_Range is new Integer range -100 .. 100;
> subtype Local_Type is Integer_Range;
>
> This is what GDB would print afterwards:
>
> (gdb) ptype local_type
> type = range 4294967196 .. 100
>
> The lower bound should be -100. The debugging info is generated
> as follow:
>
> <2><80>: Abbrev Number: 2 (DW_TAG_subrange_type)
> <81> DW_AT_lower_bound : 0xffffff9c
> <85> DW_AT_upper_bound : 100
> <86> DW_AT_name : (indirect string, offset: 0x76): homonym__get_value__local_type___XDLU_100m__100
> <8a> DW_AT_type : <0x37>
>
> And the corresponding abbrev gives the form:
>
> 2 DW_TAG_subrange_type [no children]
> DW_AT_lower_bound DW_FORM_data4
> DW_AT_upper_bound DW_FORM_data1
> DW_AT_name DW_FORM_strp
> DW_AT_type DW_FORM_ref4
> DW_AT value: 0 DW_FORM value: 0
>
> It's the dreaded DW_FORM_dataN form... And unfortunately, I get the same
> representation with pre-versions of GCC 4.9, so it looks like we're not
> going to be able to remove that bit anytime soon :-(.
Grrr and sigh. Why does GCC do that? Encoding the negative lower bound
like that actually takes up more space than simply using DW_FORM_sdata.
The comment in the GCC sources even says this is suboptimal:
/* Otherwise represent the bound as an unsigned value with the
precision of its type. The precision and signedness of the
type will be necessary to re-interpret it unambiguously. */
Luckily it seems GCC only does this when adding bounds info. But it does
seem we are stuck with it for now :{
I'll see if I can fix GCC so one day day in the far, far, future this
hack won't be necessary.
Cheers,
Mark