Hi, I've tracked the following GDB tests regression on arm-none-eabi targets to r240855. Keeping all components (gas, ld, gdb) fixed except for GCC confirmed GCC is the source of the regression. PASS->FAIL: gdb.cp/namespace.exp: print ina PASS->FAIL: gdb.cp/namespace.exp: ptype ina PASS->FAIL: gdb.cp/rtti.exp: print *e2 All three fail with the following error: No symbol "ina" in current context. (respectively e2 for the last testcase). I can see that indeed after the commit there is less debug info in namespace.cc (in particular the bits about ina is removed). Please let me know what dump would help you debug the issue. Best regards.
Ok, we're hitting /* Ugh */ which shouldn't really happen. BLOCK_VARS should be a sub-chain of gimple_bind_vars. It looks like BLOCK_VARS in this case has a common sub-chain with gimple_bind_vars instead... /* Scrap DECL_CHAIN up to BLOCK_VARS to ease GC after we no longer need gimple_bind_vars. */ tree next; tree end = NULL_TREE; if (gimple_bind_block (stmt)) end = BLOCK_VARS (gimple_bind_block (stmt)); for (tree var = gimple_bind_vars (stmt); var != end; var = next) { /* Ugh, something is violating the constraint that BLOCK_VARS is a sub-chain of gimple_bind_vars. */ if (! var) break; next = DECL_CHAIN (var); DECL_CHAIN (var) = NULL_TREE; } and we have gimple_bind_vars: c1D.2424 varD.2425 inaD.2426 yD.2460 clD.2462 BLOCK_VARS: <<< Unknown tree: imported_decl >>> <<< Unknown tree: imported_decl >>> c1D.2424 varD.2425 inaD.2426 yD.2460 clD.2462
It's that way in the original BIND_EXPR already. So it looks like we need to handle this gracefully :/ Index: gcc/gimple-low.c =================================================================== --- gcc/gimple-low.c (revision 240964) +++ gcc/gimple-low.c (working copy) @@ -420,18 +420,24 @@ lower_gimple_bind (gimple_stmt_iterator /* Scrap DECL_CHAIN up to BLOCK_VARS to ease GC after we no longer need gimple_bind_vars. */ tree next; - tree end = NULL_TREE; + /* BLOCK_VARS and gimple_bind_vars share a common sub-chain. Find + it by marking all BLOCK_VARS. */ if (gimple_bind_block (stmt)) - end = BLOCK_VARS (gimple_bind_block (stmt)); - for (tree var = gimple_bind_vars (stmt); var != end; var = next) + for (tree t = BLOCK_VARS (gimple_bind_block (stmt)); t; t = DECL_CHAIN (t)) + { + gcc_checking_assert (! TREE_VISITED (t)); + TREE_VISITED (t) = 1; + } + for (tree var = gimple_bind_vars (stmt); + var && ! TREE_VISITED (var); var = next) { - /* Ugh, something is violating the constraint that BLOCK_VARS - is a sub-chain of gimple_bind_vars. */ - if (! var) - break; next = DECL_CHAIN (var); DECL_CHAIN (var) = NULL_TREE; } + /* Unmark BLOCK_VARS. */ + if (gimple_bind_block (stmt)) + for (tree t = BLOCK_VARS (gimple_bind_block (stmt)); t; t = DECL_CHAIN (t)) + TREE_VISITED (t) = 0; lower_sequence (gimple_bind_body_ptr (stmt), data); fixes it.
Confirmed, this patch fixes the issue. Thanks!
Author: rguenth Date: Tue Oct 11 12:52:44 2016 New Revision: 240991 URL: https://gcc.gnu.org/viewcvs?rev=240991&root=gcc&view=rev Log: 2016-10-11 Richard Biener <rguenther@suse.de> PR debug/77931 * gimple-low.c (lower_gimple_bind): Handle arbitrary common sub-chains of BLOCK_VARS and gimple_bind_vars. Modified: trunk/gcc/ChangeLog trunk/gcc/gimple-low.c
Fixed.