Re: [PATCH] Always print call-clobbered registers in outer frames. (was: Re: [PATCH+DOC] Print registers not saved in the frame as "<not saved>", instead of "<optimized out>".)

[Mark Kettenis <mark dot kettenis at xs4all dot nl>]

> Date: Thu, 19 Sep 2013 20:15:20 +0100
> From: Pedro Alves <palves@redhat.com>
> 
> On 09/19/2013 05:58 PM, Pedro Alves wrote:
> 
> > Now, GDB itself could just assume that call-clobbered registers
> > are always <not saved> in frames other than the innermost, but
> > that'll be not very user friendly, I think.
> > 
> > Perhaps... we should completely toss out this patch, and go
> > the other way around.  When printing registers of an outer
> > frame, ignore DW_CFA_undefined, and read the registers
> > from the inner frame anyway...  IOW, define the values of
> > call-clobbered registers in outer frames as "the values the
> > registers would have if the inner frame returned _now_".
> > Mark, what do you think?
> > (I didn't try implementing that, but I think that'll just
> > mean ignoring the optimized_out flag when dumping registers
> > (but not when printing variables marked as living in "optimized
> > out" registers).
> > 
> 
> Like this...
> 
> --------------
> Subject: [PATCH] Always print call-clobbered registers in outer frames.
> 
> With older GCCs, when some variable lived on a call-clobbered register
> just before a call, GCC would mark such register as undefined across
> the function call, with DW_CFA_undefined.  This is interpreted by the
> debugger as meaning the variable is optimized out at that point.  That
> is, if the user does "up", and tries to print the variable.
> 
> Newer GCCs stopped doing that.  They now just don't emit a location
> for the variable, resulting in GDB printing "<optimized out>" all the
> same.  (See <https://sourceware.org/ml/gdb-patches/2012-08/msg00787.html>.)
> 
> The difference is that with binaries produced by those older GCCs, GDB
> will also print the registers themselves (info registers / p $reg) as
> "<optimized out>".  This is confusing.  See with the gdb.dwarf/dw2-op-out-param.exp
> test:
> 
> Breakpoint 2, 0x000000000040058f in breakpt ()
> (gdb) up
> #1  0x00000000004005a2 in int_param_single_reg_loc (operand0=<optimized out>, operand1=-2401054115373400575, operand2=<optimized out>)
> (gdb) info registers
> rax            0x323d7b35b0     215779849648
> rbx            0xdeadbe00deadbe01       -2401054115373400575
> rcx            <optimized out>
> rdx            0xdeadbe04deadbe05       -2401054098193531387
> rsi            <optimized out>
> rdi            <optimized out>
> rbp            0x0      0x0
> rsp            0x7fffffffda20   0x7fffffffda20
> r8             0x323d7b1f40     215779843904
> r9             0x323d00f310     215771837200
> r10            0x7fffffffd890   140737488345232
> r11            0x323d421640     215776106048
> r12            0x400430 4195376
> r13            0x7fffffffdb10   140737488345872
> r14            0x0      0
> r15            0x0      0
> rip            0x4005a2 0x4005a2 <int_param_single_reg_loc+10>
> eflags         0x202    [ IF ]
> cs             0x33     51
> ss             0x2b     43
> ds             0x0      0
> es             0x0      0
> fs             0x0      0
> gs             0x0      0
> 
> This patch makes GDB always follow this rule (which is what the
> heuristic unwinders usually do by default):
> 
>  The values of call-clobbered registers in the outer frame, if not
>  saved by the caller, are defined as being the values the registers
>  would have if the inner frame was to return immediately.
> 
> The documentation is updated to more clearly explain this.
> 
> IOW, ignore DW_CFA_undefined _when printing frame registers_, but
> not when printing variables.  This translates to, if value of a frame
> register, comes out as optimized out (that's what "not saved"
> lval_register values end up as), fetch it from the next frame.
> 
> After the patch, we get:
> 
> #0  0x000000000040058f in breakpt ()
> (gdb) info registers
> rax            0x323d7b35b0     215779849648
> rbx            0xdeadbe00deadbe01       -2401054115373400575
> rcx            0xdeadbe02deadbe03       -2401054106783465981
> rdx            0xdeadbe04deadbe05       -2401054098193531387
> rsi            0xdeadbe06deadbe07       -2401054089603596793
> rdi            0xdeadbe08deadbe09       -2401054081013662199
> rbp            0x0      0x0
> rsp            0x7fffffffda18   0x7fffffffda18
> r8             0x323d7b1f40     215779843904
> r9             0x323d00f310     215771837200
> r10            0x7fffffffd890   140737488345232
> r11            0x323d421640     215776106048
> r12            0x400430 4195376
> r13            0x7fffffffdb10   140737488345872
> r14            0x0      0
> r15            0x0      0
> rip            0x40058f 0x40058f <breakpt>
> eflags         0x202    [ IF ]
> cs             0x33     51
> ss             0x2b     43
> ds             0x0      0
> es             0x0      0
> fs             0x0      0
> gs             0x0      0
> (gdb) up
> #1  0x00000000004005a2 in int_param_single_reg_loc (operand0=<optimized out>, operand1=-2401054115373400575, operand2=<optimized out>)
> (gdb) info registers
> rax            0x323d7b35b0     215779849648
> rbx            0xdeadbe00deadbe01       -2401054115373400575
> rcx            0xdeadbe02deadbe03       -2401054106783465981
> rdx            0xdeadbe04deadbe05       -2401054098193531387
> rsi            0xdeadbe06deadbe07       -2401054089603596793
> rdi            0xdeadbe08deadbe09       -2401054081013662199
> rbp            0x0      0x0
> rsp            0x7fffffffda20   0x7fffffffda20
> r8             0x323d7b1f40     215779843904
> r9             0x323d00f310     215771837200
> r10            0x7fffffffd890   140737488345232
> r11            0x323d421640     215776106048
> r12            0x400430 4195376
> r13            0x7fffffffdb10   140737488345872
> r14            0x0      0
> r15            0x0      0
> rip            0x4005a2 0x4005a2 <int_param_single_reg_loc+10>
> eflags         0x202    [ IF ]
> cs             0x33     51
> ss             0x2b     43
> ds             0x0      0
> es             0x0      0
> fs             0x0      0
> gs             0x0      0
> (gdb)
> 
> Which is what you'd get with a newer GCC.  And, it's exactly what you
> get today if your force a return from frame #1, even with such an old
> binary, and _unpatched_ GDB:
> 
> (gdb) return
> Make breakpt return now? (y or n) y
> #0  0x00000000004005a2 in int_param_single_reg_loc (operand0=-2401054106783465981, operand1=-2401054115373400575,
>     operand2=-2401054089603596793)
> (gdb) info registers
> rax            0x323d7b35b0     215779849648
> rbx            0xdeadbe00deadbe01       -2401054115373400575
> rcx            0xdeadbe02deadbe03       -2401054106783465981
> rdx            0xdeadbe04deadbe05       -2401054098193531387
> rsi            0xdeadbe06deadbe07       -2401054089603596793
> rdi            0xdeadbe08deadbe09       -2401054081013662199
> rbp            0x0      0x0
> rsp            0x7fffffffda20   0x7fffffffda20
> r8             0x323d7b1f40     215779843904
> r9             0x323d00f310     215771837200
> r10            0x7fffffffd890   140737488345232
> r11            0x323d421640     215776106048
> r12            0x400430 4195376
> r13            0x7fffffffdb10   140737488345872
> r14            0x0      0
> r15            0x0      0
> rip            0x4005a2 0x4005a2 <int_param_single_reg_loc+10>
> eflags         0x202    [ IF ]
> cs             0x33     51
> ss             0x2b     43
> ds             0x0      0
> es             0x0      0
> fs             0x0      0
> gs             0x0      0
> (gdb)
> 
> This is rule is applied in value_of_register.  This is the function
> eval.c uses for evaluating OP_REGISTER (again, $reg, $pc, etc.), and
> related bits.  It isn't used for anything else.  The patch makes "info
> registers" and the MI equivalent use it too.  I think it just makes a
> lot of sense, as this makes it so that when printing machine registers
> ($pc, etc.), we go through a central function.
> 
> Note how gdb.mi/mi-reg-undefined.exp and
> gdb.dwarf2/dw2-reg-undefined.exp tests needed adjustment, but not
> gdb.dwarf2/dw2-op-out-param.exp, as that prints an optimized out
> variable, not register.

I strongly disagree with this change.  I think it is useful to show
that rgeisters have not been saved.  If you want to look at their
value anyway, you can always go down (or up) the stack and chase the
value of the register yourself.

I also think it doesn't make sense for the varobj subsystem to
reimplement bits of the unwinder subsystem.

If newer GCC versions are even less explicit about saving or not
saving registers than older versions, we should probably improve the
_dwarf2_frame_init_reg() implementations to indicate properly which
registers are caller-saved and which registers are callee-saved.