* [RFA] Mips, return_value_location, small structs
@ 2002-08-09 15:07 Michael Snyder
2002-08-09 17:13 ` Andrew Cagney
0 siblings, 1 reply; 9+ messages in thread
From: Michael Snyder @ 2002-08-09 15:07 UTC (permalink / raw)
To: cagney, gdb-patches
[-- Attachment #1: Type: text/plain, Size: 626 bytes --]
Re-submitted as a separate patch.
The problem: small structs returned in one (or two) registers
are aligned left in N32/N64 and aligned right in the others
(or vice versa, depending on how you cross your eyes).
There are two code paths here, one for structs of length < MIPS_REGSIZE,
and one for MIPS_REGSIZE < struct len < 2 * MIPS_REGSIZE.
The first case I've split into two paths: one for N32 &&
TYPE_CODE_STRUCT,
and a second for everything else (the later being identical to what was
there before).
In the second case, I've simply let the N32/N64 path fall thru.
The next "if" catches it and handles it correctly.
[-- Attachment #2: patch4b.diff --]
[-- Type: text/plain, Size: 2133 bytes --]
2002-08-08 Michael Snyder <msnyder@redhat.com>
* mips-tdep.c (return_value_location): Structs returned in
registers are aligned differently (n32/64 vs. o32/64).
Index: mips-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/mips-tdep.c,v
retrieving revision 1.92
diff -c -3 -p -r1.92 mips-tdep.c
*** mips-tdep.c 8 Aug 2002 00:26:51 -0000 1.92
--- mips-tdep.c 8 Aug 2002 21:48:09 -0000
*************** return_value_location (struct type *valt
*** 3608,3622 ****
if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
&& len < MIPS_SAVED_REGSIZE)
{
! /* "un-left-justify" the value in the low register */
! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
! lo->len = len;
hi->reg_offset = 0;
hi->len = 0;
}
else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
&& len > MIPS_SAVED_REGSIZE /* odd-size structs */
&& len < MIPS_SAVED_REGSIZE * 2
&& (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
TYPE_CODE (valtype) == TYPE_CODE_UNION))
{
--- 3608,3635 ----
if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
&& len < MIPS_SAVED_REGSIZE)
{
! if ((gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N32
! || gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N64)
! && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT
! || TYPE_CODE (valtype) == TYPE_CODE_UNION))
! {
! /* Values are already aligned in the low register. */
! lo->reg_offset = 0;
! }
! else
! {
! /* "un-left-justify" the value in the low register */
! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
! }
hi->reg_offset = 0;
+ lo->len = len;
hi->len = 0;
}
else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
&& len > MIPS_SAVED_REGSIZE /* odd-size structs */
&& len < MIPS_SAVED_REGSIZE * 2
+ && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N32
+ && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N64
&& (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
TYPE_CODE (valtype) == TYPE_CODE_UNION))
{
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-09 15:07 [RFA] Mips, return_value_location, small structs Michael Snyder
@ 2002-08-09 17:13 ` Andrew Cagney
2002-08-09 17:37 ` Michael Snyder
0 siblings, 1 reply; 9+ messages in thread
From: Andrew Cagney @ 2002-08-09 17:13 UTC (permalink / raw)
To: Michael Snyder; +Cc: gdb-patches
> Re-submitted as a separate patch.
>
> The problem: small structs returned in one (or two) registers
> are aligned left in N32/N64 and aligned right in the others
> (or vice versa, depending on how you cross your eyes).
I _think_ you also need to add o32. Can you check that out? Just post
the results, not the patch.
Andrew
> There are two code paths here, one for structs of length < MIPS_REGSIZE,
> and one for MIPS_REGSIZE < struct len < 2 * MIPS_REGSIZE.
>
> The first case I've split into two paths: one for N32 &&
> TYPE_CODE_STRUCT,
> and a second for everything else (the later being identical to what was
> there before).
>
> In the second case, I've simply let the N32/N64 path fall thru.
> The next "if" catches it and handles it correctly.
>
>
>
> 2002-08-08 Michael Snyder <msnyder@redhat.com>
>
> * mips-tdep.c (return_value_location): Structs returned in
> registers are aligned differently (n32/64 vs. o32/64).
>
> Index: mips-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/mips-tdep.c,v
> retrieving revision 1.92
> diff -c -3 -p -r1.92 mips-tdep.c
> *** mips-tdep.c 8 Aug 2002 00:26:51 -0000 1.92
> --- mips-tdep.c 8 Aug 2002 21:48:09 -0000
> *************** return_value_location (struct type *valt
> *** 3608,3622 ****
> if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> && len < MIPS_SAVED_REGSIZE)
> {
> ! /* "un-left-justify" the value in the low register */
> ! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
> ! lo->len = len;
> hi->reg_offset = 0;
> hi->len = 0;
> }
> else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> && len > MIPS_SAVED_REGSIZE /* odd-size structs */
> && len < MIPS_SAVED_REGSIZE * 2
> && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
> TYPE_CODE (valtype) == TYPE_CODE_UNION))
> {
> --- 3608,3635 ----
> if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> && len < MIPS_SAVED_REGSIZE)
> {
> ! if ((gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N32
> ! || gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N64)
> ! && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT
> ! || TYPE_CODE (valtype) == TYPE_CODE_UNION))
> ! {
> ! /* Values are already aligned in the low register. */
> ! lo->reg_offset = 0;
> ! }
> ! else
> ! {
> ! /* "un-left-justify" the value in the low register */
> ! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
> ! }
> hi->reg_offset = 0;
> + lo->len = len;
> hi->len = 0;
> }
> else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> && len > MIPS_SAVED_REGSIZE /* odd-size structs */
> && len < MIPS_SAVED_REGSIZE * 2
> + && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N32
> + && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N64
> && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
> TYPE_CODE (valtype) == TYPE_CODE_UNION))
> {
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-09 17:13 ` Andrew Cagney
@ 2002-08-09 17:37 ` Michael Snyder
2002-08-09 19:39 ` Andrew Cagney
0 siblings, 1 reply; 9+ messages in thread
From: Michael Snyder @ 2002-08-09 17:37 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb-patches
Andrew Cagney wrote:
>
> > Re-submitted as a separate patch.
> >
> > The problem: small structs returned in one (or two) registers
> > are aligned left in N32/N64 and aligned right in the others
> > (or vice versa, depending on how you cross your eyes).
>
> I _think_ you also need to add o32. Can you check that out? Just post
> the results, not the patch.
I know o32 needs work too, but the two can be treated separately.
I do guarantee that this patch does not make o32 any worse.
In fact, I have been testing to make sure that none of my patches
have made o32 any worse.
Can I check this one in for n32, if I promise to work on o32 next?
>
> Andrew
>
> > There are two code paths here, one for structs of length < MIPS_REGSIZE,
> > and one for MIPS_REGSIZE < struct len < 2 * MIPS_REGSIZE.
> >
> > The first case I've split into two paths: one for N32 &&
> > TYPE_CODE_STRUCT,
> > and a second for everything else (the later being identical to what was
> > there before).
> >
> > In the second case, I've simply let the N32/N64 path fall thru.
> > The next "if" catches it and handles it correctly.
> >
> >
> >
> > 2002-08-08 Michael Snyder <msnyder@redhat.com>
> >
> > * mips-tdep.c (return_value_location): Structs returned in
> > registers are aligned differently (n32/64 vs. o32/64).
> >
> > Index: mips-tdep.c
> > ===================================================================
> > RCS file: /cvs/src/src/gdb/mips-tdep.c,v
> > retrieving revision 1.92
> > diff -c -3 -p -r1.92 mips-tdep.c
> > *** mips-tdep.c 8 Aug 2002 00:26:51 -0000 1.92
> > --- mips-tdep.c 8 Aug 2002 21:48:09 -0000
> > *************** return_value_location (struct type *valt
> > *** 3608,3622 ****
> > if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> > && len < MIPS_SAVED_REGSIZE)
> > {
> > ! /* "un-left-justify" the value in the low register */
> > ! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
> > ! lo->len = len;
> > hi->reg_offset = 0;
> > hi->len = 0;
> > }
> > else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> > && len > MIPS_SAVED_REGSIZE /* odd-size structs */
> > && len < MIPS_SAVED_REGSIZE * 2
> > && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
> > TYPE_CODE (valtype) == TYPE_CODE_UNION))
> > {
> > --- 3608,3635 ----
> > if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> > && len < MIPS_SAVED_REGSIZE)
> > {
> > ! if ((gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N32
> > ! || gdbarch_tdep (current_gdbarch) -> mips_abi == MIPS_ABI_N64)
> > ! && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT
> > ! || TYPE_CODE (valtype) == TYPE_CODE_UNION))
> > ! {
> > ! /* Values are already aligned in the low register. */
> > ! lo->reg_offset = 0;
> > ! }
> > ! else
> > ! {
> > ! /* "un-left-justify" the value in the low register */
> > ! lo->reg_offset = MIPS_SAVED_REGSIZE - len;
> > ! }
> > hi->reg_offset = 0;
> > + lo->len = len;
> > hi->len = 0;
> > }
> > else if (TARGET_BYTE_ORDER == BFD_ENDIAN_BIG
> > && len > MIPS_SAVED_REGSIZE /* odd-size structs */
> > && len < MIPS_SAVED_REGSIZE * 2
> > + && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N32
> > + && gdbarch_tdep (current_gdbarch) -> mips_abi != MIPS_ABI_N64
> > && (TYPE_CODE (valtype) == TYPE_CODE_STRUCT ||
> > TYPE_CODE (valtype) == TYPE_CODE_UNION))
> > {
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-09 17:37 ` Michael Snyder
@ 2002-08-09 19:39 ` Andrew Cagney
2002-08-12 11:50 ` Michael Snyder
2002-08-14 12:23 ` Michael Snyder
0 siblings, 2 replies; 9+ messages in thread
From: Andrew Cagney @ 2002-08-09 19:39 UTC (permalink / raw)
To: Michael Snyder; +Cc: gdb-patches
> Andrew Cagney wrote:
>
>>
>
>> > Re-submitted as a separate patch.
>> >
>> > The problem: small structs returned in one (or two) registers
>> > are aligned left in N32/N64 and aligned right in the others
>> > (or vice versa, depending on how you cross your eyes).
>
>>
>> I _think_ you also need to add o32. Can you check that out? Just post
>> the results, not the patch.
>
>
> I know o32 needs work too, but the two can be treated separately.
> I do guarantee that this patch does not make o32 any worse.
> In fact, I have been testing to make sure that none of my patches
> have made o32 any worse.
>
> Can I check this one in for n32, if I promise to work on o32 next?
Why not just fix this bug for all three ABIs? o32 is just like n32/n64
in that it cross-eyes the arguments.
Andrew
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-09 19:39 ` Andrew Cagney
@ 2002-08-12 11:50 ` Michael Snyder
2002-08-14 12:23 ` Michael Snyder
1 sibling, 0 replies; 9+ messages in thread
From: Michael Snyder @ 2002-08-12 11:50 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb-patches
Andrew Cagney wrote:
>
> > Andrew Cagney wrote:
> >
> >>
> >
> >> > Re-submitted as a separate patch.
> >> >
> >> > The problem: small structs returned in one (or two) registers
> >> > are aligned left in N32/N64 and aligned right in the others
> >> > (or vice versa, depending on how you cross your eyes).
> >
> >>
> >> I _think_ you also need to add o32. Can you check that out? Just post
> >> the results, not the patch.
> >
> >
> > I know o32 needs work too, but the two can be treated separately.
> > I do guarantee that this patch does not make o32 any worse.
> > In fact, I have been testing to make sure that none of my patches
> > have made o32 any worse.
> >
> > Can I check this one in for n32, if I promise to work on o32 next?
>
> Why not just fix this bug for all three ABIs? o32 is just like n32/n64
> in that it cross-eyes the arguments.
I said I would do it. Why not accept this patch, which makes
things better than they were?
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-09 19:39 ` Andrew Cagney
2002-08-12 11:50 ` Michael Snyder
@ 2002-08-14 12:23 ` Michael Snyder
2002-08-16 10:58 ` Andrew Cagney
2002-08-19 14:39 ` Andrew Cagney
1 sibling, 2 replies; 9+ messages in thread
From: Michael Snyder @ 2002-08-14 12:23 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb-patches
Andrew Cagney wrote:
>
> > Andrew Cagney wrote:
> >
> >>
> >
> >> > Re-submitted as a separate patch.
> >> >
> >> > The problem: small structs returned in one (or two) registers
> >> > are aligned left in N32/N64 and aligned right in the others
> >> > (or vice versa, depending on how you cross your eyes).
> >
> >>
> >> I _think_ you also need to add o32. Can you check that out? Just post
> >> the results, not the patch.
> >
> >
> > I know o32 needs work too, but the two can be treated separately.
> > I do guarantee that this patch does not make o32 any worse.
> > In fact, I have been testing to make sure that none of my patches
> > have made o32 any worse.
> >
> > Can I check this one in for n32, if I promise to work on o32 next?
>
> Why not just fix this bug for all three ABIs? o32 is just like n32/n64
> in that it cross-eyes the arguments.
Andrew, o32 works as it is. I can't find anything related to this
patch that still needs to be done for o32. Unles you can suggest
something specific, may I check this patch in?
Michael
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-14 12:23 ` Michael Snyder
@ 2002-08-16 10:58 ` Andrew Cagney
2002-08-19 14:39 ` Andrew Cagney
1 sibling, 0 replies; 9+ messages in thread
From: Andrew Cagney @ 2002-08-16 10:58 UTC (permalink / raw)
To: Michael Snyder; +Cc: gdb-patches
Michael,
Having fixed a few build problems (did you encounter similar?) I'm
finding that I can't get an n32 GDB to recognize an n32 binary. Vis:
mips-sgi-irix6.5/gdb/gdb": not in executable format: File format not
recognized
file ./gdb/gdb
./gdb/gdb: ELF N32 MSB mips-4 dynamic executable (not stripped)
MIPS - version 1
Is there anything special I need?
Andrew
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-14 12:23 ` Michael Snyder
2002-08-16 10:58 ` Andrew Cagney
@ 2002-08-19 14:39 ` Andrew Cagney
2002-08-20 11:49 ` Michael Snyder
1 sibling, 1 reply; 9+ messages in thread
From: Andrew Cagney @ 2002-08-19 14:39 UTC (permalink / raw)
To: Michael Snyder; +Cc: gdb-patches
> I know o32 needs work too, but the two can be treated separately.
>> > I do guarantee that this patch does not make o32 any worse.
>> > In fact, I have been testing to make sure that none of my patches
>> > have made o32 any worse.
>> >
>> > Can I check this one in for n32, if I promise to work on o32 next?
>
>>
>> Why not just fix this bug for all three ABIs? o32 is just like n32/n64
>> in that it cross-eyes the arguments.
>
>
> Andrew, o32 works as it is. I can't find anything related to this
> patch that still needs to be done for o32. Unles you can suggest
> something specific, may I check this patch in?
FYI, this patch should now be redundant. I rewrote the n32/n64
extract/store return value code.
Andrew
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [RFA] Mips, return_value_location, small structs
2002-08-19 14:39 ` Andrew Cagney
@ 2002-08-20 11:49 ` Michael Snyder
0 siblings, 0 replies; 9+ messages in thread
From: Michael Snyder @ 2002-08-20 11:49 UTC (permalink / raw)
To: Andrew Cagney, gdb-patches
Andrew Cagney wrote:
>
> > I know o32 needs work too, but the two can be treated separately.
> >> > I do guarantee that this patch does not make o32 any worse.
> >> > In fact, I have been testing to make sure that none of my patches
> >> > have made o32 any worse.
> >> >
> >> > Can I check this one in for n32, if I promise to work on o32 next?
> >
> >>
> >> Why not just fix this bug for all three ABIs? o32 is just like n32/n64
> >> in that it cross-eyes the arguments.
> >
> >
> > Andrew, o32 works as it is. I can't find anything related to this
> > patch that still needs to be done for o32. Unles you can suggest
> > something specific, may I check this patch in?
>
> FYI, this patch should now be redundant. I rewrote the n32/n64
> extract/store return value code.
Check, this patch is no longer pertainant.
Withdrawn.
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2002-08-20 18:49 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-08-09 15:07 [RFA] Mips, return_value_location, small structs Michael Snyder
2002-08-09 17:13 ` Andrew Cagney
2002-08-09 17:37 ` Michael Snyder
2002-08-09 19:39 ` Andrew Cagney
2002-08-12 11:50 ` Michael Snyder
2002-08-14 12:23 ` Michael Snyder
2002-08-16 10:58 ` Andrew Cagney
2002-08-19 14:39 ` Andrew Cagney
2002-08-20 11:49 ` Michael Snyder
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox