Commit Message
On Fri, 29 May 2015 15:43:19 +0200, Yao Qi wrote:
> Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> > The terminology seems bogus there.
> >
> > "partially ambiguous" was meant the chain:
> > main -> a -> <???> -> d
> > An intersection of all possible chains.
>
> Sounds like "partially ambiguous" is equivalent to "ambiguous".
Yes, probably, I am not sure how to call it all myself.
> If that is right, the assert below is too strict, isn't?
Yes, it is too strict, this is why I agree with the fix by Andreas.
> /* See call_site_find_chain_1 why there is no way to reach the bottom callee
> PC again. In such case there must be two different code paths to reach
> it, therefore some of the former determined intermediate PCs must differ
> and the unambiguous chain gets shortened. */
> gdb_assert (result->callers + result->callees < result->length);
>
> > but that doe snot matter). Consequently its elements from the middle are
> > being removed and there remains only some few unambiguous top and
> > bottom ones.
>
> If there is no call sites removed from the chain during the intersection,
> CALLERS + CALLEES == LENGTH, right?
Just I expected there always has to be some site removed from the chain.
I do not find obvious it does not have to. But maybe someone else finds it
obvious.
> in function chain_candidate,
> result->length is set by the length of a chain. If this chain is the
> shortest one, CALLERS + CALLEES == LENGTH otherwise,
> CALLERS + CALLEES < LENGTH. Is it right?
It is right now. But when one does not think about self-tail-calls then even
the shortest one will get one frame removed.
> If so, we need to relax the
> condition in the assert and update the comments.
Yes, attached with updated comment.
> > I did not realize that there can be self-tail-call:
> > main(0x100) -> a(0x200) -> d(0x400)
> > main(0x100) -> a(0x280) -> a(0x200) -> d(0x400)
> > which intersects to:
> > main(0x100) -> <???>? -> a(0x200) -> d(0x400)
> > And so if the first chain was chosen the
> > main(0x100) -> a(0x200) -> d(0x400)
> > then the final intersection has callers+callees==length.
>
> What are the definitions of CALLERS, CALLEES, top and bottom? given this example?
top=CALLERS=main(0x100), therefore 1
bottom=CALLEES=d(0x400), therefore 1
top = topmost, where you can go by GDB "up" commands, also called "prev" in
struct frame_info.
bottom = bottommost, where you can go by GDB "down" commands, also called
"next" in struct frame_info.
Jan
2015-05-29 Andreas Schwab <schwab@linux-m68k.org>
Jan Kratochvil <jan.kratochvil@redhat.com>
PR symtab/18392
* dwarf2-frame-tailcall.c (pretended_chain_levels): Correct
assertion.
* dwarf2loc.c (chain_candidate): Likewise.
Comments
Jan Kratochvil <jan.kratochvil@redhat.com> writes:
>> > I did not realize that there can be self-tail-call:
>> > main(0x100) -> a(0x200) -> d(0x400)
>> > main(0x100) -> a(0x280) -> a(0x200) -> d(0x400)
>> > which intersects to:
>> > main(0x100) -> <???>? -> a(0x200) -> d(0x400)
>> > And so if the first chain was chosen the
>> > main(0x100) -> a(0x200) -> d(0x400)
>> > then the final intersection has callers+callees==length.
>>
>> What are the definitions of CALLERS, CALLEES, top and bottom? given
>> this example?
>
> top=CALLERS=main(0x100), therefore 1
> bottom=CALLEES=d(0x400), therefore 1
>
> top = topmost, where you can go by GDB "up" commands, also called "prev" in
> struct frame_info.
>
> bottom = bottommost, where you can go by GDB "down" commands, also called
> "next" in struct frame_info.
OK, I understand what does top/bottom mean. Since they are numeric
values, what does these number mean? for example, if CALLERS is 3 and
CALLEES is 2, what does the chain look like?
The code change in the patch looks reasonable to me, but comments change
doesn't, probably because I don't fully understand it. I'll take a
deeper look next Monday.
On Fri, 29 May 2015 18:33:01 +0200, Yao Qi wrote:
> OK, I understand what does top/bottom mean. Since they are numeric
> values, what does these number mean?
CALLERS and CALLEES together with LENGTH say what data is at what indexes of
CALL_SITE:
struct call_site_chain
{
/* Initially CALLERS == CALLEES == LENGTH. For partially ambiguous result
CALLERS + CALLEES < LENGTH. */
int callers, callees, length;
/* Variably sized array with LENGTH elements. Later [0..CALLERS-1] contain
top (GDB "prev") sites and [LENGTH-CALLEES..LENGTH-1] contain bottom
(GDB "next") sites. One is interested primarily in the PC field. */
struct call_site *call_site[1];
};
> for example, if CALLERS is 3 and
> CALLEES is 2, what does the chain look like?
main(0x100) -> x(0x150) -> y(0x200) -> <???>? -> a(0x200) -> d(0x400)
And if LENGTH is 7 then:
call_site[0] = main(0x100)
call_site[1] = x(0x150)
call_site[2] = y(0x200)
call_site[3] = garbage
call_site[4] = garbage
call_site[5] = a(0x200)
call_site[6] = d(0x400)
Thanks,
Jan
Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> diff --git a/gdb/dwarf2loc.c b/gdb/dwarf2loc.c
> index 3aa8ddd..68d6cb4 100644
> --- a/gdb/dwarf2loc.c
> +++ b/gdb/dwarf2loc.c
> @@ -825,9 +825,9 @@ chain_candidate (struct gdbarch *gdbarch, struct call_site_chain **resultp,
>
> /* See call_site_find_chain_1 why there is no way to reach the bottom callee
> PC again. In such case there must be two different code paths to reach
> - it, therefore some of the former determined intermediate PCs must differ
> - and the unambiguous chain gets shortened. */
> - gdb_assert (result->callers + result->callees < result->length);
> + it. Still it may CALLERS+CALLEES==LENGTH in the case of optional
> + tail-call calling itself. */
> + gdb_assert (result->callers + result->callees <= result->length);
I am not a native English speaker, but I can't parse the comment. How
about "CALLERS + CALLEES equal to LENGTH in the case of self tail-call"?
Otherwise, the patch is OK to me.
On Mon, 01 Jun 2015 13:35:02 +0200, Yao Qi wrote:
> I am not a native English speaker, but I can't parse the comment. How
> about "CALLERS + CALLEES equal to LENGTH in the case of self tail-call"?
Used your comment part.
> Otherwise, the patch is OK to me.
Checked in:
e0619de699ae6e86d8b93fa96a7668aef2e9636a
Thanks,
Jan
@@ -197,7 +197,7 @@ pretended_chain_levels (struct call_site_chain *chain)
return chain->length;
chain_levels = chain->callers + chain->callees;
- gdb_assert (chain_levels < chain->length);
+ gdb_assert (chain_levels <= chain->length);
return chain_levels;
}
@@ -825,9 +825,9 @@ chain_candidate (struct gdbarch *gdbarch, struct call_site_chain **resultp,
/* See call_site_find_chain_1 why there is no way to reach the bottom callee
PC again. In such case there must be two different code paths to reach
- it, therefore some of the former determined intermediate PCs must differ
- and the unambiguous chain gets shortened. */
- gdb_assert (result->callers + result->callees < result->length);
+ it. Still it may CALLERS+CALLEES==LENGTH in the case of optional
+ tail-call calling itself. */
+ gdb_assert (result->callers + result->callees <= result->length);
}
/* Create and return call_site_chain for CALLER_PC and CALLEE_PC. All the