https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84379
Bug ID: 84379
Summary: Problems with sol2.c GTY handling
Product: gcc
Version: 8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: other
Assignee: unassigned at gcc dot gnu.org
Reporter: ro at gcc dot gnu.org
Target Milestone: ---
Target: *-*-solaris2.*
While investigating the jit testsuite failures on Solaris with the native as
(cf. PR jit/84288), I noticed that there's no GTY markup in gcc/config/sol2.c.
Trying to fix that (adding GTY(()) markers to struct comdat_entry and
solaris_comdat_htab and including gt-sol2.h resulted in code that wouldn't even
compile:
In file included from /vol/gcc/src/hg/trunk/solaris/gcc/coretypes.h:423:0,
from /vol/gcc/src/hg/trunk/solaris/gcc/config/sol2.c:23:
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h: In instantiation of 'void
gt_ggc_mx(hash_table<E>*) [with E = comdat_entry_hasher]':
./gt-sol2.h:28:19: required from here
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h:1049:39: error: 'ggc_maybe_mx'
is not a member of 'comdat_entry_hasher'
E::ggc_maybe_mx (h->m_entries[i]);
^
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h: In instantiation of 'void
gt_pch_nx(hash_table<E>*) [with D = comdat_entry_hasher]':
./gt-sol2.h:44:19: required from here
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h:1084:33: error: 'pch_nx' is not
a member of 'comdat_entry_hasher'
D::pch_nx (h->m_entries[i]);
^
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h: In instantiation of 'void
hashtab_entry_note_pointers(void*, void*, gt_pointer_operator, void*) [with D =
comdat_entry_hasher; gt_pointer_operator = void (*)(void*, void*)]':
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h:1076:74: required from 'void
gt_pch_nx(hash_table<E>*) [with D = comdat_entry_hasher]'
./gt-sol2.h:44:19: required from here
/vol/gcc/src/hg/trunk/solaris/gcc/hash-table.h:1067:47: error: 'pch_nx' is not
a member of 'comdat_entry_hasher'
D::pch_nx (map->m_entries[i], op, cookie);
^
make[3]: *** [/vol/gcc/src/hg/trunk/solaris/gcc/config/t-sol2:36: sol2.o] Error
1
I haven't yet been able to fix that.
I've also tried a build with --enable-checking=yes,gcac to determine if the
jit testsuite failures observed are really caused by GC issues. However, the
32-bit build would abort early with an OOM error, and the 64-bit build times
were so over-the-top that they probably wouldn't complete in a week.
