[Bug c/42688] New: gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[Bug c/42688] New: gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[mnemo at minimum dot se]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1294 bytes --]

When you build sqlite using GCC and "-O2" parameter you get the following
warning. During runtime "EXPR_FULLSIZE-nSize" is mostly non-zero so it seems
that the GCC optimizer makes some assumptions that doesn't not always hold
true.

In function ‘memset’,
    inlined from ‘exprDup’ at sqlite3.c:61011:
/usr/include/bits/string3.h:82: warning: call to ‘__warn_memset_zero_len’
declared with attribute warning: memset used with constant zero length
parameter; this could be due to transposed parameters

Repro steps:
0. install GCC 4.3.3
1. fetch and unpack sqlite from this URL:
http://www.sqlite.org/sqlite-amalgamation-3.6.22.tar.gz
2. run "gcc -O2 sqlite3.c"

Background: this issue was originally posted to the sqlite project as a sqlite
bug but they dismissed it as a GCC bug instead (which I believe is correct):
http://www.sqlite.org/src/tktview?name=f51d950180


-- 
           Summary: gcc 4.3.3 with -O2 thinks a often non-zero expression is
                    always zero
           Product: gcc
           Version: 4.3.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: mnemo at minimum dot se


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug c/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2010-01-12 09:10 -------
Please attach preprocessed source of the offending file.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug c/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[mnemo at minimum dot se]

------- Comment #2 from mnemo at minimum dot se  2010-01-17 15:34 -------
Using gcc 4.4.1, I see this:
1. "gcc -O2 sqlite3.c" repros the bug
2. "gcc -E sqlite3.c > pp.c" followed by "gcc -O2 pp.c" does _NOT_ repro the
bug

However, still using gcc 4.4.1, if I do:
1. "gcc -O2 -E sqlite3.c > ppo.c" followed by "gcc -O2 ppo.c" does repro the
bug

Finally, I also tried to upgrade to Ubuntu Lucid today which gave me gcc 4.4.3
and there I can't get the bug to repro at all.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug c/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[mnemo at minimum dot se]

------- Comment #3 from mnemo at minimum dot se  2010-01-17 15:39 -------
I tried to attach the output of "gcc -E -O2 sqlite3.c" but it's 1.5MB and
bugzilla rejects anything bigger than 1KB. Even the unmodified source is larger
than this limit, even though you can download it via the sqlite.org URL I
posted.

Since this doesn't repro on gcc 4.4.3, maybe the bug should be closed though?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug middle-end/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2010-02-07 20:04 -------
You can gzip the preprocessed source if needed.  Also It might be the case we
are jump threading in one case and not in the other which causes the zero
expression to show up and that causes glibc's headers to produce the warning
...


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |middle-end


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug middle-end/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[mnemo at minimum dot se]

------- Comment #5 from mnemo at minimum dot se  2010-02-07 21:34 -------
The gzipped output is also too big for this bugzilla instance to handle. It
should be easy for you to download the *.tar.gz from the description and just
run "gcc -E -O2" on it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

[Bug middle-end/42688] gcc 4.3.3 with -O2 thinks a often non-zero expression is always zero

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42688

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |INVALID

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-19 02:20:20 UTC ---
No testcase provided in almost 2 years so closing as invalid.