CC: Warning: -O5 overwrites previously set optimization level of -O3
I'm trying to get a clean compile of the Boost 1.32.0 library using the Studio Express August build, and in the course of that (this post is not about Boost), I get the following:
sunpro-C++-action bin/boost/libs/filesystem/build/libboost_filesystem.a/sunpro/release/exception. o
/opt/SPROexpress/opt/SUNWspro//bin//CC -O -xlang=c99 -mt \
-library=stlport4 -features=tmplife -features=tmplrefstatic -errtags=yes \
-D_POSIX_THREAD_SEMANTICS -c -DNDEBUG -DNDEBUG -fast -O4 -xtarget=generic \
-library=stlport4 -features=tmplife -features=tmplrefstatic -erroff=wbadinitl,wbadasgl \
-Qoption ccfe -complextmplexp \
-I"bin/boost/libs/filesystem/build" -I"/thirdparty/lib/boost_1_32_0" -I"/thirdparty/lib/boost_1_32_0" \
-o "bin/boost/libs/filesystem/build/libboost_filesystem.a/sunpro/release/exception .o" \
"/thirdparty/lib/boost_1_32_0/libs/filesystem/build/../src/exception.cpp"
CC: Warning: -O5 overwrites previously set optimization level of -O3
Maybe I'm doing something wrong in the way that both -O and -O4 get specified on the command line, but why is the error message specifying -O5 and -O3, neither of which I specified?
[1204 byte] By [
herteg] at [2007-11-26 10:10:40]
# 1
I see what's going on here. -O expands to -O3. That is overridden by the -O5 which is included in the expansion of -fast.
Four things about this:
(1) The warning message should say "overrides", not "overwrites".
(2) There is no corresponding warning when the -O5 is overridden by the following -O4. Or is it?
(3) The effect of multiple -Olevel options is unspecified; that should be corrected in the manual.
(4) The Sun Studio 11 C++ User's Guide is inconsistent on the meaning of the -O option:
A.2.61 -O
The -O macro now expands to -xO3 instead of -xO2.
A.2.145 -xOlevel
On the SPARC Platform:
The -O option is equivalent to the -xO2 option.
# 3
Herteg,
> I see what's going on here. -O expands to -O3. That
> is overridden by the -O5 which is included in the
> expansion of -fast.
This is correct.
> Four things about this:
>
> (1) The warning message should say "overrides", not
> "overwrites".
I think that there should be no warning at all. We have "rightmost wins" rule for the compiler flags, and both -O and -fast are documented well enough to understand which -O<n> (or -xO<n>) they will be translated into.
The problem is that I don't see this "rightmost wins" in the documentation for -O<n>/-xO<n>.
> (2) There is no corresponding warning when the -O5 is
> overridden by the following -O4. Or is it?
I believe there shouldn't be one in the first place.
> (3) The effect of multiple -Olevel options is
> unspecified; that should be corrected in the manual.
Well, it is specified, it's just not documented :-)
> (4) The Sun Studio 11 C++ User's Guide is
> inconsistent on the meaning of the -O option:
>
> A.2.61 -O
> The -O macro now expands to -xO3 instead of -xO2.
>
> A.2.145 -xOlevel
> On the SPARC Platform:
> The -O option is equivalent to the -xO2 option.
The second one is wrong. But I suppose you already know this.
Thank you very much for bringing this to our attention! Would you please file a bug through the web interface: http://bugs.sun.com/services/bugreport/index.jsp ?
Thanks,
Boris
# 4
>> (1) The warning message should say "overrides", not "overwrites".
> I think that there should be no warning at all. We have "rightmost wins" rule for the compiler flags, and both -O and -fast are documented well enough to understand which -O<n> (or -xO<n>) they will be translated into.
Actually, I agree on wanting no warning, unless it were off by default and there were some kind of "tell me all the things I normally want suppressed" flag. I just didn't dare hope.
> The problem is that I don't see this "rightmost wins" in the documentation for -O<n>/-xO<n>.
You're right. I encountered the same problem with -erroff=... . Definitely this kind of blanket rule belongs in the docs.
