[Bug target/106056] New: Missing call to targetm.asm_out.final_postscan_insn after processing an asm_input

[Bug target/106056] New: Missing call to targetm.asm_out.final_postscan_insn after processing an asm_input

[piannetta at kalrayinc dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106056

            Bug ID: 106056
           Summary: Missing call to targetm.asm_out.final_postscan_insn
                    after processing an asm_input
           Product: gcc
           Version: 10.3.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: piannetta at kalrayinc dot com
  Target Milestone: ---

Created attachment 53190
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53190&action=edit
patch which adds a call to final_postscan_insn after the processing of an
asm_input

When an instruction is output, there should be a call to `final_postscan_insn`
but this is not the case after the processing of an asm_input.

I did not check from which version this call is missing but the surrounding
lines where last edited around 2007.

This patch adds the missing call to `final_postscan_insn`.

[Bug middle-end/106056] Missing call to targetm.asm_out.final_postscan_insn after processing an asm_input

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106056

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-06-27
             Status|UNCONFIRMED                 |WAITING
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
Thanks for the suggestion, do you have a test-case which is affected?

[Bug middle-end/106056] Missing call to targetm.asm_out.final_postscan_insn after processing an asm_input

[piannetta at kalrayinc dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106056

--- Comment #2 from Paul Iannetta <piannetta at kalrayinc dot com> ---
Right now, I can't pinpoint a test-case within the official gcc testsuite, and
the list of targets which define TARGET_ASM_FINAL_POSTSCAN_INSN is limited to:
mips, avr and m68k.  From my understanding the only target that might be
affected is avr.

However, we maintain a VLIW target and rely on this hook to append our bundle
separator and there is a discrepancy between the treatment of extended asm and
basic asm blocks.

```
asm ("nop":::);
asm ("nop":::);
```

Each extended asm block triggers final_postscan and our bundle separator is
rightfully inserted.  However,

```
asm ("nop");
asm ("nop");
```

does not trigger final_postscan and the bundle separator is missing.