[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 Eric Botcazou changed: What|Removed |Added Attachment #44066|0 |1 is obsolete|| --- Comment #13 from Eric Botcazou --- Created attachment 44081 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=44081&action=edit Tentative fix v2 To be tested.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 --- Comment #12 from Eric Botcazou --- > I wonder why this doesn't seem to happen with DWARF-2 exception handling. OK, the difference is that dw2_build_landing_pads does: lp->landing_pad = gen_label_rtx (); emit_label (lp->landing_pad); LABEL_PRESERVE_P (lp->landing_pad) = 1; whereas sjlj_emit_dispatch_table does: lp->landing_pad = dispatch_label; In other words, there is a landing pad per post-landing pad with DWARF-2 and a common landing pad for all post-landing pads with SJLJ so, when you redirect the edges from post-landing to landing pads, you lose the 1->1 mapping with SJLJ.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 --- Comment #11 from xantares09 at hotmail dot com --- Oh, ok
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 --- Comment #10 from Eric Botcazou --- > Yes, I applied your patch, now the build fails with another error: This one is yours, you need to compile the 8.1 cross with the 8.1 native.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 --- Comment #9 from xantares09 at hotmail dot com --- Yes, I applied your patch, now the build fails with another error: gcc -c -I./ -I/usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/adalib/../adainclude -I/usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/adalib -I. -I/home/xantares/projects/aur-scripts/mingw-w64-gcc/src/gcc/gcc/ada -g -O2 -W -Wall -gnatpg -gnata -I- /home/xantares/projects/aur-scripts/mingw-w64-gcc/src/gcc-8.1.0/gcc/ada/fmap.adb fmap.adb:304:18: "Null_FD" is undefined gnatmake: "/home/xantares/projects/aur-scripts/mingw-w64-gcc/src/gcc-8.1.0/gcc/ada/fmap.adb" compilation error
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 --- Comment #8 from Eric Botcazou --- Another (checking) assertion triggers in fix_up_crossing_landing_pad: rtx_insn *insn = BB_END (e->src); rtx note = find_reg_note (insn, REG_EH_REGION, NULL_RTX); gcc_assert (note != NULL); gcc_checking_assert (INTVAL (XEXP (note, 0)) == old_lp->index); Since 2 elements of cfun->eh->lp_array can share the same landing pad, you can have a different REG_EH_REGION note on one of the edges. I wonder why this doesn't seem to happen with DWARF-2 exception handling.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 Eric Botcazou changed: What|Removed |Added Attachment #44060|0 |1 is obsolete|| Attachment #44061|0 |1 is obsolete|| --- Comment #7 from Eric Botcazou --- Created attachment 44066 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=44066&action=edit Tentative fix To be tested.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638
--- Comment #6 from Eric Botcazou ---
The key point is that several elements of cfun->eh->lp_array can share the same
landing pad:
(gdb) p *cfun->eh->lp_array->m_vecdata[1]
$11 = {next_lp = 0x0, region = 0x7695c898,
post_landing_pad = 0x7695df80, landing_pad = 0x76a6b200, index = 1}
(gdb) p *cfun->eh->lp_array->m_vecdata[2]
$12 = {next_lp = 0x0, region = 0x7695c8f0,
post_landing_pad = 0x7695de80, landing_pad = 0x76a6b200, index = 2}
(gdb) p *cfun->eh->lp_array->m_vecdata[1]->region
$13 = {outer = 0x0, inner = 0x0, next_peer = 0x0, index = 1, type = ERT_TRY,
u = {eh_try = {first_catch = 0x76964398, last_catch = 0x76964398},
allowed = {type_list = 0x76964398, label = 0x76964398,
filter = 0}, must_not_throw = {failure_decl = 0x76964398,
failure_loc = 4137042840}}, landing_pads = 0x769648c0,
exc_ptr_reg = 0x76a6a2a0, filter_reg = 0x76a6a150,
use_cxa_end_cleanup = false}
(gdb) p *cfun->eh->lp_array->m_vecdata[2]->region
$14 = {outer = 0x0, inner = 0x0, next_peer = 0x7695c898, index = 2,
type = ERT_CLEANUP, u = {eh_try = {first_catch = 0x0, last_catch = 0x0},
allowed = {type_list = 0x0, label = 0x0, filter = 0}, must_not_throw = {
failure_decl = 0x0, failure_loc = 0}}, landing_pads = 0x769643c0,
exc_ptr_reg = 0x76a6a8e8, filter_reg = 0x0, use_cxa_end_cleanup = false}
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638
--- Comment #5 from Eric Botcazou ---
It's a direct fallout of the fix for PR rtl-optimization/85393:
FOR_EACH_EDGE (e, ei, bb->preds)
{
gcc_assert (e->flags & EDGE_EH);
if (BB_PARTITION (bb) == BB_PARTITION (e->src))
all_diff = false;
else
all_same = false;
}
which creates incoming edges to landing pads that are not EH edges.
[Bug ada/85638] [8/9 regression] build failure for Ada runtime with SJLJ exceptions on x86
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85638 Eric Botcazou changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |ebotcazou at gcc dot gnu.org --- Comment #4 from Eric Botcazou --- Investigating.
