bug-binutils
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug gas/16288] New: Specifying an ISA in a .set directive causes a warn


From: hp at sourceware dot org
Subject: [Bug gas/16288] New: Specifying an ISA in a .set directive causes a warning if "lower" than set from command-line
Date: Tue, 03 Dec 2013 02:20:32 +0000

http://sourceware.org/bugzilla/show_bug.cgi?id=16288

            Bug ID: 16288
           Summary: Specifying an ISA in a .set directive causes a warning
                    if "lower" than set from command-line
           Product: binutils
           Version: 2.25 (HEAD)
            Status: NEW
          Severity: normal
          Priority: P2
         Component: gas
          Assignee: unassigned at sourceware dot org
          Reporter: hp at sourceware dot org
            Target: mips*-*

Seen with recent master of binutils, for example 310bf259c3524c2954; from
code-inspection likely to be present for 2.24 as well.

The following construct is common with in-lined code (seen in alsa-lib and
userspace-rcu), but gets a warning if the setting is "lower" than one specified
on the command-line.

For a mips-elf target and a file s.s, with the contents:

 .set push
 .set mips2
 nop
 .set pop

Observe:
$ gas/as-new s.s
$ gas/as-new -march=34kc s.s
s.s: Assembler messages:
s.s:2: Warning: the `dsp' extension requires MIPS32 revision 2 or greater
s.s:2: Warning: the `mt' extension requires MIPS32 revision 2 or greater

Note that the warning comes from the ".set mips2" directive, regardless of
whether it's followed by code; the test-case can actually be reduced to that
single line.  The -march option comes from gcc when specifying the same option
or from the configuration.

The warning is bogus; the .set implementation seems to miss clearing
command-line-specified ASEs when specifying an ISA.  I believe that is how it's
intuitively should work; this warning is a regression from, like, 8 months ago.

-- 
You are receiving this mail because:
You are on the CC list for the bug.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]