This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Fix complex float on sparc
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: davem at davemloft dot net
- Cc: gdb-patches at sourceware dot org
- Date: Wed, 28 Sep 2011 11:39:28 +0200 (CEST)
- Subject: Re: [PATCH] Fix complex float on sparc
- References: <20110928.012027.2236138577869370946.davem@davemloft.net>
> Date: Wed, 28 Sep 2011 01:20:27 -0400 (EDT)
> From: David Miller <davem@davemloft.net>
>
> Lots of testcase failures have crept into the sparc targets,
> I'll try to fix as many as I can.
>
> This gets the complex float cases of callfuncs.exp passing
> again for both 32-bit and 64-bit.
>
> For sparc32 it's pass in memory, return in float regs.
>
> For sparc64 it's pass in float regs (<= 16 bytes) or memory (> 16
> bytes), always return in float regs.
>
> Ok to commit?
Looks good. Just a small request inline...
> gdb/
>
> 2011-09-27 David S. Miller <davem@davemloft.net>
>
> * sparc-tdep.h (SPARC_F2_REGNUM, SPARC_F3_REGNUM, SPARC_F4_REGNUM,
> SPARC_F5_REGNUM, SPARC_F6_REGNUM, SPARC_F7_REGNUM): New enums.
> * sparc-tdep.c (sparc_complex_floating_p): New function.
> (sparc32_store_arguments): Handle complex floats.
> (sparc32_extract_return_value): Likewise.
> (sparc32_store_return_value): Likewise.
> (sparc32_stabs_argument_has_addr): Likewise.
> * sparc64-tdep.c (sparc64_complex_floating_p): New function.
> (sparc64_store_floating_fields): Handle complex floats.
> (sparc64_store_arguments): Likewise.
> (sparc64_store_return_value): Likewise.
>
> diff --git a/gdb/sparc-tdep.c b/gdb/sparc-tdep.c
> index faa7b3a..8541a31 100644
> --- a/gdb/sparc-tdep.c
> +++ b/gdb/sparc-tdep.c
> @@ -1238,12 +1258,25 @@ sparc32_extract_return_value (struct type *type, struct regcache *regcache,
> gdb_assert (!sparc_structure_or_union_p (type));
> gdb_assert (!(sparc_floating_p (type) && len == 16));
>
> - if (sparc_floating_p (type))
> + if (sparc_floating_p (type)
> + || sparc_complex_floating_p (type))
I would have put both conditions on the same line, since this is
wasting a bit of vertical space. Don't see this list growing in the
near future and I don't think there's a significant difference in
readability that way.
> @@ -1281,13 +1314,26 @@ sparc32_store_return_value (struct type *type, struct regcache *regcache,
> gdb_assert (!(sparc_floating_p (type) && len == 16));
> gdb_assert (len <= 8);
>
> - if (sparc_floating_p (type))
> + if (sparc_floating_p (type)
> + || sparc_complex_floating_p (type))
Same here.
> diff --git a/gdb/sparc64-tdep.c b/gdb/sparc64-tdep.c
> index 0430ecf..097f658 100644
> --- a/gdb/sparc64-tdep.c
> +++ b/gdb/sparc64-tdep.c
> @@ -622,11 +642,13 @@ static void
> sparc64_store_floating_fields (struct regcache *regcache, struct type *type,
> const gdb_byte *valbuf, int element, int bitpos)
> {
> + int len = TYPE_LENGTH (type);
> +
> gdb_assert (element < 16);
>
> - if (sparc64_floating_p (type))
> + if (sparc64_floating_p (type)
> + || (sparc64_complex_floating_p (type) && len <= 16))
Here.
> @@ -886,7 +908,8 @@ sparc64_store_arguments (struct regcache *regcache, int nargs,
> if (element < 16)
> sparc64_store_floating_fields (regcache, type, valbuf, element, 0);
> }
> - else if (sparc64_floating_p (type))
> + else if (sparc64_floating_p (type)
> + || sparc64_complex_floating_p (type))
Here.
> @@ -1067,7 +1090,8 @@ sparc64_store_return_value (struct type *type, struct regcache *regcache,
> if (TYPE_CODE (type) != TYPE_CODE_UNION)
> sparc64_store_floating_fields (regcache, type, buf, 0, 0);
> }
> - else if (sparc64_floating_p (type))
> + else if (sparc64_floating_p (type)
> + || sparc64_complex_floating_p (type))
And here.
Would appreciate it if you could change this.