make install fails due to "-Wno-overlength-strings"

make install fails due to "-Wno-overlength-strings"

[Todd Sayers]

I'm attempting to install gcc 4.3.6 on RHEL 5.5 x86_64 using the installed (default, from Redhat) gcc version 4.1.2.  Everything [except the `make install`] appears to work well.  Only the `make install` of gcc 4.3.6 fails.  It dies with this error:

cc1: error: unrecognized command line option "-Wno-overlength-strings"

gmp,mpfr, and mpc all compiled without error, passed `make check`, and were installed via `make install.`  There were no issues.  GCC 4.3.6 compiles without error (make) and passes almost 96,000 tests before `make check` dies (after >2hrs using -j4).  GCC 4.3.6 was configured 
with:

../gcc-4.3.6/configure --prefix=/foo/HPC_apps/AMD --with-gmp-lib=/foo/HPC_apps/AMD/lib --with-gmp-include=/foo/HPC_apps/AMD/include --with-mpfr-include=/foo/HPC_apps/AMD/include --with-mpfr-lib=/foo/HPC_apps/AMD/lib

Please note that (per the instructions) I am NOT building in the source directory.  I Googled until I reached the end of the Internet... and, well, after that I decided to write this e-mail.  What gives?  Is there an environment variable that can be set to force gcc to not attempt to use this option (in the configure logs, it discovers that "-Wno-overlength-strings" is not an available option.  WTF?
 		 	   		  

Re: make install fails due to "-Wno-overlength-strings"

[Ian Lance Taylor]

Todd Sayers <tscoltrane@hotmail.com> writes:

> I'm attempting to install gcc 4.3.6 on RHEL 5.5 x86_64 using the installed (default, from Redhat) gcc version 4.1.2.  Everything [except the `make install`] appears to work well.  Only the `make install` of gcc 4.3.6 fails.  It dies with this error:
>
> cc1: error: unrecognized command line option "-Wno-overlength-strings"
>
> gmp,mpfr, and mpc all compiled without error, passed `make check`, and were installed via `make install.`  There were no issues.  GCC 4.3.6 compiles without error (make) and passes almost 96,000 tests before `make check` dies (after >2hrs using -j4).  GCC 4.3.6 was configured 
> with:
>
> ../gcc-4.3.6/configure --prefix=/foo/HPC_apps/AMD --with-gmp-lib=/foo/HPC_apps/AMD/lib --with-gmp-include=/foo/HPC_apps/AMD/include --with-mpfr-include=/foo/HPC_apps/AMD/include --with-mpfr-lib=/foo/HPC_apps/AMD/lib
>
> Please note that (per the instructions) I am NOT building in the source directory.  I Googled until I reached the end of the Internet... and, well, after that I decided to write this e-mail.  What gives?  Is there an environment variable that can be set to force gcc to not attempt to use this option (in the configure logs, it discovers that "-Wno-overlength-strings" is not an available option.  WTF?
>  		 	   		  

It sounds like you are saying that you get the error about
-Wno-overlength-strings when you run "make install".  But running "make
install" should not actually compile anything.  So something has gone
wrong.  Perhaps your timestamps are not right, who knows.

I would recommend running "make" a couple of times until you are sure
that nothing is being recompiled.  Then run "make install".  If it still
fails, show us the last 100 lines or so of make output.

Ian

Re: make install fails due to "-Wno-overlength-strings"

[Ryan Hill]

[-- Attachment #1: Type: text/plain, Size: 2176 bytes --]

On Thu, 3 Nov 2011 10:34:23 -0500
Todd Sayers <tscoltrane@hotmail.com> wrote:

> 
> I'm attempting to install gcc 4.3.6 on RHEL 5.5 x86_64 using the installed (default, from Redhat) gcc version 4.1.2.  Everything [except the `make install`] appears to work well.  Only the `make install` of gcc 4.3.6 fails.  It dies with this error:
> 
> cc1: error: unrecognized command line option "-Wno-overlength-strings"
> 
> gmp,mpfr, and mpc all compiled without error, passed `make check`, and were installed via `make install.`  There were no issues.  GCC 4.3.6 compiles without error (make) and passes almost 96,000 tests before `make check` dies (after >2hrs using -j4).  GCC 4.3.6 was configured 
> with:
> 
> ../gcc-4.3.6/configure --prefix=/foo/HPC_apps/AMD --with-gmp-lib=/foo/HPC_apps/AMD/lib --with-gmp-include=/foo/HPC_apps/AMD/include --with-mpfr-include=/foo/HPC_apps/AMD/include --with-mpfr-lib=/foo/HPC_apps/AMD/lib
> 
> Please note that (per the instructions) I am NOT building in the source directory.  I Googled until I reached the end of the Internet... and, well, after that I decided to write this e-mail.  What gives?  Is there an environment variable that can be set to force gcc to not attempt to use this option (in the configure logs, it discovers that "-Wno-overlength-strings" is not an available option.  WTF?

We've run into this several times with 4.3.  Never did find a solution and
ended up going away with 4.4.

https://bugs.gentoo.org/241604

Comment 13 (which i sadly just noticed now): 
Since this is the "make install" stage, there shouldn't be any recompilation
necessary, but in my log file, it shows that "Must remake target
`cc1-checksum.o'", which caused the compile error mentioned in the opening. The
interesting bit is that inside the work/build/gcc dir there is a
"cc1-checksum.c" file, but no .o nor executable was created for it (but
cc1plus-checksum.o does exist).


--
fonts, gcc-porting,                  it makes no sense how it makes no sense
toolchain, wxwidgets                           but i'll take it free anytime
@ gentoo.org                EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

RE: make install fails due to "-Wno-overlength-strings"

[Todd Sayers]

Ryan et al,

     Thanks for adding another data point what I've 
begun to suspect (not "know," but "suspect").  I gather [from the 
"interWebs"] that this problem is specific to gcc 
4.1.2 compiling gcc 4.3.6 on 64-bit.  Whatever.  My solution was to 
complete the make-install with "-i" for the first version.  Then, I 
compiled a clean 4.3.6 using the "dirty" 4.3.6.  Lastly, use the "clean"
 4.3.6 to compile the "final/good/shiny" 4.3.6.  Recursive error 
correction or just silly?  Meh -- all DejaGNU tests pass... and no 
errors were reported.  I guess that's the best that I can think up.  
Your thoughts?

Thanks to Ian, too.  I appreciate the assistance. 
 I ran through this dozens of times  before regressing/devolving/etc to 
my "recursive error correction" method.  I didn't see anything in the 
last 200 lines of the make log that was meaningful -- beyond what I 
posted (`tail -[100,200]|less`is my standard query).

I hope that 
this is a "valid" approach.  I used the "clean gcc 4.3.6" to do a 
subsequent error-free build of gcc 4.3.6.  A little knuckle-head, 
sure... but valid, right?  Or have I just worked the errors into the 
dough?

 - Todd