[Bug lto/48467] New: [LTO/driver] Anomalous behavior of -save-temps

[Bug lto/48467] New: [LTO/driver] Anomalous behavior of -save-temps

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48467

           Summary: [LTO/driver] Anomalous behavior of -save-temps
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: minor
          Priority: P3
         Component: lto
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: d.g.gorbachev@gmail.com


Created attachment 23884
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23884
DIFF file with 4 tests

1. After running the 1st test, notice the absence of one generated assembler
file (1.s), which is deleted by lto1. In toplev.c,

1851  if (flag_wpa)
1852    unlink_if_ordinary (asm_file_name);

2. As the 2nd and 3rd tests show, it can also be misused as an output file for
lto1, so its original, written by cc1, content gets lost.

3. The 4th test gives good, expected results.

[Bug lto/48467] [LTO/driver] Anomalous behavior of -save-temps

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48467

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2011.04.06 10:01:51
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-06 10:01:51 UTC ---
Confirmed.

What we could do is try to arrange for the WPA lto1 stage to receive
-o /dev/null, as it will not produce regular output.  Or to simply
avoid calling init_asm_output () for -fwpa.  So, sth like

Index: toplev.c
===================================================================
--- toplev.c    (revision 172023)
+++ toplev.c    (working copy)
@@ -1743,7 +1743,8 @@ lang_dependent_init (const char *name)
     return 0;
   input_location = save_loc;

-  init_asm_output (name);
+  if (!flag_wpa)
+    init_asm_output (name);

   /* If stack usage information is desired, open the output file.  */
   if (flag_stack_usage)
@@ -1848,8 +1849,6 @@ finalize (bool no_backend)
        fatal_error ("error writing to %s: %m", asm_file_name);
       if (fclose (asm_out_file) != 0)
        fatal_error ("error closing %s: %m", asm_file_name);
-      if (flag_wpa)
-       unlink_if_ordinary (asm_file_name);
     }

   if (stack_usage_file)


which likely needs fixups throughout the compiler in case of
missing asm_out_file checks.

[Bug lto/48467] [LTO/driver] Anomalous behavior of -save-temps

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48467

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-08 13:10:31 UTC ---
Author: rguenth
Date: Fri Apr  8 13:10:27 2011
New Revision: 172181

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=172181
Log:
2011-04-08  Richard Guenther  <rguenther@suse.de>

    PR lto/48467
    * toplev.c (lang_dependent_init): Do not open asm_out_file
    in WPA mode, nor perform debug machinery initialization.
    (finalize): Do not unlink asm_out_file in WPA mode.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/toplev.c

[Bug lto/48467] [LTO/driver] Anomalous behavior of -save-temps

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48467

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.7.0

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-08 13:10:51 UTC ---
Fixed for 4.7.

[Bug lto/48467] [LTO] Anomalous behavior of -save-temps

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48467

Dmitry Gorbachev <d.g.gorbachev at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[LTO/driver] Anomalous      |[LTO] Anomalous behavior of
                   |behavior of -save-temps     |-save-temps

--- Comment #4 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> 2011-04-19 18:53:51 UTC ---
Second bug is now bug 48691.