Thanks, Olav.
So, it seems like I have a compiler bug that is specific to SPARC, the
cause of my zero-sized modules was:
source=Index2Page.cpp object=Index2Page.o libtool=no DEPDIR=.deps
depmode=none
/bin/bash ../../depcomp CC -DHAVE_CONFIG_H -D_REENTRANT -D_PTHREADS
-DFALCONDB -DSTORAGE_ENGINE -DNAMESPACE=Nfs -I. -I../../include
-I../../include -I../../regex -I../../sql -I./TransformLib -I../../zlib
-g0 -xO4 -mt -m64 -xtarget=generic -fsimple=1 -xbuiltin=%all -xlibmil
-xlibmopt -noex -DHAVE_RWLOCK_T -Wno-invalid-offsetof -features=except
-c -o Index2Page.o Index2Page.cpp
CC: Warning: Option -Wno-invalid-offsetof passed to ld, if ld is
invoked, ignored otherwise
"MemoryManager.h", line 88: Warning (Anachronism): The prior declaration
for operator new(unsigned long) has an exception specification.
"MemoryManager.h", line 100: Warning (Anachronism): The prior
declaration for operator new[](unsigned long) has an exception
specification.
"HdrState.h", line 23: Warning: Identifier expected instead of "}".
"HdrState.h", line 29: Warning: Identifier expected instead of "}".
"Index2Page.cpp", line 601: Warning: pageNumber hides Page::pageNumber.
"Index2Page.cpp", line 633: Warning: pageNumber hides Page::pageNumber.
"Index2Page.cpp", line 663: Warning: pageNumber hides Page::pageNumber.
"Index2Page.cpp", line 723: Warning: pageNumber hides Page::pageNumber.
8 Warning(s) detected.
CODEGEN IR TREE SYNTAX ERROR AT LINE
WITHIN ROUTINE __1cKIndex2PageIvalidate6MpnDDbb_pnKValidation_pnGBitmap_i_v_
! (Index2Page.cpp:598)
<0>T[51] int cbranch T[48] labels: (L[190] if (0x0,0x1) 1) (L[191] if
(0x0,0x0) 0) attr:
<0> T[48] uchar lt L[33]=(0x0,0x1c) 28 T[47] attr:
<71> L[33] "" int CONST (0x0,0x1c) 28
<994> T[47] longlong conv T[46] attr: (no_overflow=1)
<385> T[46] short ifetch T[45] attr: can_access: L[0]
<370> T[45] PTR short size 8 align 8 plus
L[176]=%v341 L[30]=(0x0,0x1a) 26 attr: (no_overflow=1)
<3> L[176] "" PTR strty size 8 align
8 VAR %v341 dbg_sym_id: {123,1}
<71> L[30] "" int CONST (0x0,0x1a) 26
<0> T[49] int labelref L[190]=(0x0,0x496fa13) 77003283
L[9]=(0x0,0x1) 1 attr:
<0> L[190] "" int CONST (0x0,0x496fa13) 77003283
Repeating the same "decomp CC" using Sun Studio 11 completed successfully.
This has been logged as CR 6808316 against Sun Studio 12.
Regards,
Tim
Olav Sandstaa wrote:
> Hi Tim,
>
> [sending this as a second email since this refers to the internal
> pushbuild testing]
>
> We also compile and run every revision that is checked into the MySQL
> repository on a set of test machines. Two of these runs Nevada build 101
> and I can not see that there has been any problems building Falcon on
> these platforms recently.
>
> You can see the results from building and running mysql-6.0 on
> "Pushbuild" by going to the following page:
>
>
> http://clustra.norway.sun.com/~bteam/pb2/web.py?branch=7&template=show_pushes
>
>
> If you look at the second "big table" (which starts at about the middle
> of the page and the first word is "Push") you can look up the column
> named "snv101a x86 max". If you click on the second "symbol" (just after
> the disk symbol) you can get a text page that shows how "pushbuild"
> configured and compiled MySQL with Falcon on Nevada (and if the
> background of the symbol is green the compilation should be OK).
>
> In this table each row corresponds to one "push" (or check-in) into the
> mysql-6.0 tree.
>
> Olav
>
>
> Tim Cook - PAE wrote:
>> [Apologies if this is the wrong audience]
>>
>> A quick check of bugs & list archive comes up empty for me on this.
>>
>> I am attempting to build mysql-6.0, revno 2718 (pretty recent),
>> --with-falcon, on Solaris Nevada build 106, using Sun Studio 12.
>>
>> It fails when linking mysqld due to missing symbols referenced from a
>> module in libfalcon.a, and my triage so far shows up two empty objects:
>>
>> ukiah[bash]$ pwd
>> /u/build/mysql-6.0_revno2718/build.sparcv9/storage/falcon
>> ukiah[bash]$ ls -log Index{,2}Page.o
>> -rw-r--r-- 1 0 Feb 19 15:53 Index2Page.o
>> -rw-r--r-- 1 0 Feb 19 15:53 IndexPage.o
>>
>> If anyone already knows what issue I am looking at, please let me
>> know. Otherwise, I will keep digging tomorrow.
>>
>> Regards,
>> Tim
>>
>>
>
--
Tim Cook
Performance and Applications Engineering
Menlo Park, CA USA
<> Sun Microsystems
Ph: +1 650 257 4709
Ext: 69841
