https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #8 from Tomáš Glozar ---
It looks like somehow a value RTX with rt_cselib set to NULL gets into the
hashmap:
(gdb) f 1
#1 rtx_equal_for_cselib_1 (x=0x2674608, y=0x26747f8,
memmode=memmode@entry=E_VOIDmode, depth=depth@entry=1)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #7 from Frank Scheiner ---
Hi again,
this problem was bisected by me and I identified the following commit as first
bad commit:
[5fc4d3e1837](https://gcc.gnu.org/git/?p=gcc.git;a=blobdiff;f=gcc/cselib.cc;h=2abc763a3f8ed4f7bb2460b8f
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #6 from Frank Scheiner ---
Dear Richard,
would it be helpful to bisect this problem or is it already known where and
when the problem was introduced?
I anyhow wanted to look into how those cross-compilers over at [1] are built
and
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #5 from Frank Scheiner ---
(In reply to Richard Biener from comment #2)
> Confirmed on trunk - we ICE in cselib during var-tracking, so
> -fno-var-tracking would be a workaround (or -g0).
Also confirming that both - i.e. using `-fno
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #4 from Frank Scheiner ---
(In reply to Richard Biener from comment #3)
Hi Richard,
in case you wanted me to test this reduced test case, I ran it through as if it
was the file in question. I needed to remove `-Werror=strict-protot
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #3 from Richard Biener ---
Reduced testcase:
enum { REFCOUNT_ADD_NOT_ZERO_OVF } refcount_warn_saturate();
int refcount_inc_not_zero_old, fib_info_num_path___trans_tmp_11,
rtnh_ok_remaining, fib_create_info_nexthop_nh_0, fib_crea
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
Richard Biener changed:
What|Removed |Added
Keywords||ice-on-valid-code
Status|U
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111425
--- Comment #1 from Frank Scheiner ---
Created attachment 55905
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55905&action=edit
the gzip compressed preprocessed file
Added the preprocessed file (gzip compressed due to size of 4.4 MiB).