ruby-core@ruby-lang.org archive (unofficial mirror)
 help / color / mirror / Atom feed
From: "eightbitraptor (Matthew Valentine-House)" <noreply@ruby-lang.org>
To: ruby-core@ruby-lang.org
Subject: [ruby-core:110147] [Ruby master Bug#19028] GCC12 Introduces new warn flags `-Wuse-after-free`
Date: Thu, 29 Sep 2022 12:50:46 +0000 (UTC)	[thread overview]
Message-ID: <redmine.journal-99404.20220929125045.31800@ruby-lang.org> (raw)
In-Reply-To: redmine.issue-19028.20220928171211.31800@ruby-lang.org

Issue #19028 has been updated by eightbitraptor (Matthew Valentine-House).


nobu (Nobuyoshi Nakada) wrote in #note-4:
> GCC 12 on GitHub Actions doesn’t show that warning.
> https://github.com/ruby/ruby/actions/runs/3146751201/jobs/5115549022#step:13:130
> 
> I’m not sure what differs.

mame (Yusuke Endoh) wrote in #note-5:
> I am not able to repro with `gcc-12 (Ubuntu 12.1.0-2ubuntu1~22.04) 12.1.0` and x86_64-linux.

Apologies. I had a look into this and realised that this warning only occurs when `optflags=-O0`.

With `O0` I can successfully repro this on:

* `gcc version 12.2.1 20220819 (Red Hat 12.2.1-2) (GCC)` on `Fedora 36` - where I originally saw this warning.
* `gcc version 12.1.0 (Ubuntu 12.1.0-2ubuntu1~22.04)`
* `gcc version 12.2.0 (Ubuntu 12.2.0-3ubuntu1)` on `Ubuntu Kinetic Kudu (development branch) 22.10`

I don't know if we should justify lowering a warning level globally for a warning that only appears in unoptimized code - what do you think?


----------------------------------------
Bug #19028: GCC12 Introduces new warn flags `-Wuse-after-free`
https://bugs.ruby-lang.org/issues/19028#change-99404

* Author: eightbitraptor (Matthew Valentine-House)
* Status: Open
* Priority: Normal
* Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN
----------------------------------------
[GCC 12 introduced a new warning flag `-Wuse-after-free`](https://github.com/gcc-mirror/gcc/commit/671a283636de75f7ed638ee6b01ed2d44361b8b6) which attempts to warn about uses of pointers to dynamically allocated objects that have been rendered indeterminate by a call to a deallocation function

Details of the levels are in [the C++ Dialect Options section of the GCC documentation](https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html).

Compiling with `-Wall` uses the default setting of `2` for this flag. Which warns on the `TRY_WITH_GC` macro defined in `gc.c`

```
gc.c: In function ‘objspace_xrealloc’:                                                                             
gc.c:12213:33: warning: pointer ‘ptr’ may be used after ‘realloc’ [-Wuse-after-free]                               
12213 |     TRY_WITH_GC(new_size, mem = realloc(ptr, new_size));                                                   
      |                                 ^~~~~~~~~~~~~~~~~~~~~~                                                     
gc.c:12123:19: note: in definition of macro ‘TRY_WITH_GC’                                                          
12123 |         else if ((expr)) {                                   \                                             
      |                   ^~~~                                                                                     
In file included from ./include/ruby/defines.h:72,                                                                 
                 from ./include/ruby/ruby.h:25,                                                                    
                 from constant.h:13,                                                                               
                 from gc.c:97:                                                                                     
gc.c:12213:33: note: call to ‘realloc’ here                                                                        
12213 |     TRY_WITH_GC(new_size, mem = realloc(ptr, new_size));                                                   
      |                                 ^~~~~~~~~~~~~~~~~~~~~~                                                                                                                                                                        
./include/ruby/backward/2/assume.h:43:46: note: in definition of macro ‘RB_LIKELY’                                 
   43 | # define RB_LIKELY(x)   (__builtin_expect(!!(x), 1))                                                       
      |                                              ^                                                             
gc.c:12116:13: note: in expansion of macro ‘LIKELY’ 
12116 |         if (LIKELY((expr))) {                                \                                             
      |             ^~~~~~                       
gc.c:12213:5: note: in expansion of macro ‘TRY_WITH_GC’
12213 |     TRY_WITH_GC(new_size, mem = realloc(ptr, new_size));                                                   
      |     ^~~~~~~~~~~            
```

My understanding is that if `realloc` returns a null pointer then the memory requested for reallocation is guaranteed to not be touched (according to [the Open Group](https://pubs.opengroup.org/onlinepubs/9699919799/functions/realloc.html) - thank you @nobu for bringing this to my attention).

Given that this is a new warning, my proposed solution is to lower the level down to the base level `1`. This will only warn on unconditional calls to deallocation functions or successful calls to realloc.

I've opened a PR that sets `-Wuse-after-free=1` only for GCC versions > 11. An alternative approach might be to use `#pragma GCC diagnostic` to suppress this just for GCC but I opted for what I thought was the easiest fix to start with.

[Github PR #6465](https://github.com/ruby/ruby/pull/6465)



-- 
https://bugs.ruby-lang.org/

  parent reply	other threads:[~2022-09-29 12:50 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-28 17:12 [ruby-core:110133] [Ruby master Bug#19028] GCC12 Introduces new warn flags `-Wuse-after-free` eightbitraptor (Matthew Valentine-House)
2022-09-28 18:48 ` [ruby-core:110137] " mame (Yusuke Endoh)
2022-09-28 23:52 ` [ruby-core:110143] " nobu (Nobuyoshi Nakada)
2022-09-29  1:42 ` [ruby-core:110144] " mame (Yusuke Endoh)
2022-09-29 12:50 ` eightbitraptor (Matthew Valentine-House) [this message]
2022-09-30  6:34 ` [ruby-core:110152] " nobu (Nobuyoshi Nakada)
2022-09-30  8:00 ` [ruby-core:110153] " eightbitraptor (Matthew Valentine-House)
2022-10-04  2:12 ` [ruby-core:110175] " nobu (Nobuyoshi Nakada)

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-list from there: mbox

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

  List information: https://www.ruby-lang.org/en/community/mailing-lists/

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

  git send-email \
    --in-reply-to=redmine.journal-99404.20220929125045.31800@ruby-lang.org \
    --to=ruby-core@ruby-lang.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).