[Bug lto/88081] [7/8/9 Regression] ICE in lto_varpool_replace_node, at lto/lto-symtab.c:109

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88081

--- Comment #1 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
The checks that we do replace definition by non-defition in symbol merging.
=Would be possible to bisect this? GCC 8 produces:

_ZTVN10__cxxabiv117__class_type_infoE/1 (_ZTVN10__cxxabiv117__class_type_infoE)
@0x7ffff69e1180
  Type: variable                                                                
  Body removed by symtab_remove_unreachable_nodes                               
  Visibility: externally_visible no_reorder undef external public weak comdat
comdat_group:_ZTVN10__cxxabiv117__class_type_infoE one_only virtual artificial
  next sharing asm name: 12                                                     
  References:                                                                   
  Referring: _ZTI1b/4 (addr)                                                    
  Read from file: 1.o                                                           
  Availability: not-ready                                                       
  Varpool flags: initialized read-only const-value-known                        

_ZTVN10__cxxabiv117__class_type_infoE/12
(_ZTVN10__cxxabiv117__class_type_infoE) @0x7ffff69e1380
  Type: variable definition analyzed                                            
  Visibility: externally_visible undef external public weak comdat
comdat_group:_ZTVN10__cxxabiv117__class_type_infoE one_only virtual artificial
  previous sharing asm name: 1                                                  
  References: _ZTIN10__cxxabiv117__class_type_infoE/14 (addr)_ZN1a1bEv/18
(addr)
  Referring: _ZTI1c/15 (addr)                                                   
  Read from file: 2.o                                                           
  Availability: not-ready                                                       
  Varpool flags: initialized read-only const-value-known                        

and has resolution info

2                                                                               
1.o 6                                                                           
221 4a2d2cb92dbf08ab PREVAILING_DEF_IRONLY _ZNK1b1cEv                           
235 4a2d2cb92dbf08ab PREVAILING_DEF_IRONLY _ZTS1b                               
251 4a2d2cb92dbf08ab PREVAILING_DEF_IRONLY _ZTI1b                               
275 4a2d2cb92dbf08ab PREVAILING_DEF_IRONLY _ZTV1b                               
201 4a2d2cb92dbf08ab UNDEF __gxx_personality_v0                                 
270 4a2d2cb92dbf08ab UNDEF _ZTVN10__cxxabiv117__class_type_infoE                
2.o 5                                                                           
230 6b69d9d40039828d PREVAILING_DEF_IRONLY _ZNK1c5m_fn2Ev                       
259 6b69d9d40039828d PREVAILING_DEF_IRONLY _ZTS1c                               
274 6b69d9d40039828d PREVAILING_DEF_IRONLY _ZTI1c                               
210 6b69d9d40039828d PREVAILING_DEF_IRONLY _ZTV1c                               
282 6b69d9d40039828d UNDEF _ZTVN10__cxxabiv117__class_type_infoE                

same happens for trunk but it happens to work.
so it may be just wrong handling of UNDEF symbol in the symtab.