https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83372
David Malcolm <dmalcolm at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |GC, ice-on-valid-code,
| |needs-bisection,
| |needs-reduction, patch
Status|ASSIGNED |WAITING
See Also| |https://gcc.gnu.org/bugzill
| |a/show_bug.cgi?id=84263
Summary|Compiler segfaults building |ICE in GC within
|Mir |gt_ggc_mx<deferred_access_c
| |heck> building Mir
--- Comment #15 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
I tried Nathan's patch from here:
[PR c++/84263] GC ICE with decltype
https://gcc.gnu.org/ml/gcc-patches/2018-02/msg00435.html
With the patch, it successfully compiles the file.
Without the patch: it ICEs, with corrupted memory in the buffer pointer to by
"checks":
(gdb) p *x
$10 = {value = 0x7ffe4295b7e0, checks = 0x7ffe425ca6c0, qualifying_scope = 0x0}
(gdb) p *x->checks
$14 = {m_vecpfx = {m_alloc = 1919901535, m_using_auto_storage = 0, m_num =
1953709151}, m_vecdata = {{binfo = 0x75665f73693a3a64,
decl = 0x763c6e6f6974636e, diag_decl = 0x26262a282064696f, loc =
1803036713}}}
(gdb) call debug(x->value)
full-name "class std::unique_ptr<xkb_state, void (*)(xkb_state*)>"
needs-constructor needs-destructor X() has-type-conversion X(constX&)
this=(X&) n_parents=0 use_template=1 interface-unknown
(gdb) call inform (x->location, "token is here")
/builddir/build/BUILD/mir-5500595810c28c150a3bd9edf19b392c2aeab932/src/server/frontend/wayland/wayland_connector.cpp:980:17:
note: token is here
Hence it looks like this is a duplicate of PR c++/84263 (but hard to be sure
without properly reducing it).
Nathan: does this look like a duplicate to you? Any ideas on how to reduce the
reproducer? (the ~1hr time to hit the bug when forcing GC is making this
awkward).
Alan: are you able to test Nathan's bug to see if it fixes the issue for you?
Thanks.
