public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug lto/105399] [12 Regression] -O2/-Ofast -flto ICEs as internal compiler error: verify_cgraph_node failed
Date: Wed, 27 Apr 2022 09:03:35 +0000	[thread overview]
Message-ID: <bug-105399-4-umTL07qSYf@http.gcc.gnu.org/bugzilla/> (raw)
In-Reply-To: <bug-105399-4@http.gcc.gnu.org/bugzilla/>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105399

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Because of -Ofast initial compilation, opt_for_fn (decl,
flag_semantic_interposition) is false for all the decls for which a cgraph node
is created, but they have NULL DECL_FUNCTION_SPECIFIC_OPTIMIZATION, and thus
they look up the bit in optimization_default_node.
That results in node->semantic_interposition to be 0 as well.
Next comes free_lang_data_in_decl, which does:

  if (TREE_CODE (decl) == FUNCTION_DECL)
...
      if (gimple_has_body_p (decl) || (node && node->thunk))
...
          if (!DECL_FUNCTION_SPECIFIC_OPTIMIZATION (decl))
            DECL_FUNCTION_SPECIFIC_OPTIMIZATION (decl)
              = optimization_default_node;
Now, free_lang_data_in_decl is called on 3 FUNCTION_DECLs,
_ZN1SC1Ev aka __ct_comp, which is node->alias && !node->semantic_interposition,
gimple_has_body_p (decl) is false for it and we don't do anything.
Next it is called on _ZN1SC4Ev aka __ct, which has NULL node and
gimple_has_body_p (decl) is false for it too.
Last it is called for _ZN1SC2Ev aka __ct_base, which is !node->alias &&
!node->semantic_interposition, gimple_has_body_p (decl) is true and we set
DECL_FUNCTION_SPECIFIC_OPTIMIZATION (decl) to optimization_default_node.
Then we stream all this out, and stream it back, except that
optimization_default_node is one for -O2 with x_flag_semantic_interposition set
for it.
_ZN1SC2Ev passes the verification, node->semantic_interposition is false
and opt_for_fn (decl, flag_semantic_interposition) is false as well.
But for _ZN1SC1Ev we ICE on:
  if (definition && externally_visible
      && semantic_interposition
         != opt_for_fn (decl, flag_semantic_interposition))
    {
      error ("semantic interposition mismatch");
      error_found = true;
    }
That is because for aliases which don't have a body, we kept
DECL_FUNCTION_SPECIFIC_OPTIMIZATION (decl) NULL.
One way to fix this would be
-  if (definition && externally_visible
+  if (definition
+      && externally_visible
+      && (!alias || thunk)
       && semantic_interposition
          != opt_for_fn (decl, flag_semantic_interposition))
     {
       error ("semantic interposition mismatch");
       error_found = true;
     }
i.e. don't bother verification of mismatches for aliases.  Another would be
to adjust DECL_FUNCTION_SPECIFIC_OPTIMIZATION (decl) in free_lang_data_in_decl
even for aliases.

  parent reply	other threads:[~2022-04-27  9:03 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-04-26 22:27 [Bug lto/105399] New: " slyfox at gcc dot gnu.org
2022-04-27  6:10 ` [Bug lto/105399] " rguenth at gcc dot gnu.org
2022-04-27  8:49 ` marxin at gcc dot gnu.org
2022-04-27  9:03 ` jakub at gcc dot gnu.org [this message]
2022-04-27  9:08 ` jakub at gcc dot gnu.org
2022-04-27  9:57 ` jakub at gcc dot gnu.org
2022-04-28 13:46 ` cvs-commit at gcc dot gnu.org
2022-04-28 13:49 ` jakub at gcc dot gnu.org

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bug-105399-4-umTL07qSYf@http.gcc.gnu.org/bugzilla/ \
    --to=gcc-bugzilla@gcc.gnu.org \
    --cc=gcc-bugs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).