[lfs-support] gcc 4.9.0 - update

Frans de Boer frans at fransdb.nl
Fri May 2 12:45:01 PDT 2014


On 04/30/2014 10:42 PM, Frans de Boer wrote:
> Since the day gcc-4.9.0 came out I am trying to get gcc compiled without
> problems. That is working, by running the tests is a different story.
> Where gcc-4.8.2 passes the tests, gcc-4.9.0 is failing - without
> exception - every time on the next tests:
>
> --------------------------------------------------------------------------
> Running /sources-bss/gcc-4.9.0/gcc/testsuite/gcc.dg/tsan/tsan.exp ...
> FAIL: c-c++-common/tsan/thread_leak1.c  -O0  output pattern test, is
> ==================
> WARNING: ThreadSanitizer: thread leak (pid=18366)
>    Thread T1 (tid=18371, finished) created by main thread at:
>      #0 pthread_create <null>:0 (libtsan.so.0+0x00000004b0d3)
>      #1 main
> /sources-bss/gcc-4.9.0/gcc/testsuite/c-c++-common/tsan/thread_leak1.c:12
> (thread_leak1.exe+0x000000000a90)
>
> SUMMARY: ThreadSanitizer: thread leak ??:0 pthread_create
> ==================
> ThreadSanitizer: reported 1 warnings
> , should match WARNING: ThreadSanitizer: thread leak.*(
> |
> )SUMMARY: ThreadSanitizer: thread leak.*main.*(
> |
> )
> FAIL: c-c++-common/tsan/thread_leak1.c  -O2  output pattern test, is
> ==================
> WARNING: ThreadSanitizer: thread leak (pid=18384)
>    Thread T1 (tid=18389, finished) created by main thread at:
>      #0 pthread_create <null>:0 (libtsan.so.0+0x00000004b0d3)
>      #1 main
> /sources-bss/gcc-4.9.0/gcc/testsuite/c-c++-common/tsan/thread_leak1.c:12
> (thread_leak1.exe+0x000000000942)
>
> SUMMARY: ThreadSanitizer: thread leak ??:0 pthread_create
> ==================
> ThreadSanitizer: reported 1 warnings
> , should match WARNING: ThreadSanitizer: thread leak.*(
> |
> )SUMMARY: ThreadSanitizer: thread leak.*main.*(
> |
> )
> Running /sources-bss/gcc-4.9.0/gcc/testsuite/gcc.dg/ubsan/ubsan.exp ...
> --------------------------------------------------------------------------
>
> and again with g++
>
> --------------------------------------------------------------------------
> Running /sources-bss/gcc-4.9.0/gcc/testsuite/g++.dg/tsan/tsan.exp ...
> FAIL: c-c++-common/tsan/thread_leak1.c  -O0  output pattern test, is
> ==================
> WARNING: ThreadSanitizer: thread leak (pid=8094)
>    Thread T1 (tid=8099, finished) created by main thread at:
>      #0 pthread_create <null>:0 (libtsan.so.0+0x00000004b0d3)
>      #1 main
> /sources-bss/gcc-4.9.0/gcc/testsuite/c-c++-common/tsan/thread_leak1.c:12
> (thread_leak1.exe+0x000000000a70)
>
> SUMMARY: ThreadSanitizer: thread leak ??:0 pthread_create
> ==================
> ThreadSanitizer: reported 1 warnings
> , should match WARNING: ThreadSanitizer: thread leak.*(
> |
> )SUMMARY: ThreadSanitizer: thread leak.*main.*(
> |
> )
> FAIL: c-c++-common/tsan/thread_leak1.c  -O2  output pattern test, is
> ==================
> WARNING: ThreadSanitizer: thread leak (pid=8112)
>    Thread T1 (tid=8117, finished) created by main thread at:
>      #0 pthread_create <null>:0 (libtsan.so.0+0x00000004b0d3)
>      #1 main
> /sources-bss/gcc-4.9.0/gcc/testsuite/c-c++-common/tsan/thread_leak1.c:12
> (thread_leak1.exe+0x000000000922)
>
> SUMMARY: ThreadSanitizer: thread leak ??:0 pthread_create
> ==================
> ThreadSanitizer: reported 1 warnings
> , should match WARNING: ThreadSanitizer: thread leak.*(
> |
> )SUMMARY: ThreadSanitizer: thread leak.*main.*(
> |
> )
> Running /sources-bss/gcc-4.9.0/gcc/testsuite/g++.dg/ubsan/ubsan.exp ...
> --------------------------------------------------------------------------
>
> I am using a AMD64, 8GB system.
> I have rebuild chapter 5 several times over, to the exact letter as in
> the SVN docs. I did the same for chapter 6, up to gcc-4.9.0.
> The report Bruce gave (24th) did not mentioned any failure in the gcc
> tests. So, the question is what is going wrong?
>
> Frans

I did extensive testing these last few days and I noticed every time 
that when compiling/testing with '-O[1-3]' or no optimization given, the 
above errors are the only ones.
However, when I compile/test the BSS products with '-O0' (except for 
glibc which can't be compiled without optimization) more errors appear - 
like lto, ABI related and internal compiler errors - where there where 
none when using gcc-4.8.2.

How come that I get the feeling that gcc-4.9.0 is not yet ripe for 
production use?

I also wonder what is different on my machine, compared with other 
machines where things seem to be running without any error. I have build 
a standard TC and the BSS is build without automatic scripts to rule out 
any typo or interpretation of some sort, still yielding the same result 
as with the scripted builds. I could build all without strange errors 
when using gcc-4.8.2, the trouble started with the introduction of 
gcc-4.9.0.

Maybe wait a few months for gcc-4.9.1?

Frans.



More information about the lfs-support mailing list