Warning on AIX

[siddus]

Hi

My dev environment is -

AIX mars 1 5 005A28AD4C00
Thread model: aix
gcc version 3.3.2

I am trying to build a popular open source cryptography library (crypto++ http://www.cryptopp.com/#download v5.5.2).
However, I get a warning message while compiling a CPP file (dll.cpp)

misc.h: In member function `const T& CryptoPP::Singleton<T, F, instance>::Ref()
const [with T = CryptoPP::DL_KeyAgreementAlgorithm_DH<CryptoPP::Integer,
CryptoPP::NoCofactorMultiplication>, F =
CryptoPP::NewObject<CryptoPP::DL_KeyAgreementAlgorithm_DH<CryptoPP::Integer,
CryptoPP::NoCofactorMultiplication> >, int instance = 0]':
dh.h:85: instantiated from `const CryptoPP::DL_KeyAgreementAlgorithm<typename GROUP_PARAMETERS::Element>& CryptoPP::DH_Domain<GROUP_PARAMETERS, COFACTOR_OPTION>::GetKeyAgreementAlgorithm() const [with GROUP_PARAMETERS = CryptoPP::DL_GroupParameters_GFP_DefaultSafePrime, COFACTOR_OPTION = CryptoPP::NoCofactorMultiplication]'
dh.h:92: instantiated from here
misc.h:121: warning: sorry: semantics of inline function static data `char
s_objectState' are wrong (you'll wind up with multiple copies)

The reference to misc.h (line 121) is in several places and I get many warning messages like that. I am going bonkers trying to figure out what this means and how to get rid of this.

Line 120/121 looks like this -

120: static simple_ptr<T> s_pObject;
121: static char s_objectState = 0;

I have attached the files in question and will be glad if someone can help me with this.

Thanks

Cheers!

sid

[siddus]

I tried removing the static part from the variable initialization on line 121 and tried compiling the file using g++ -DNDEBUG -g -O2 -pipe -c dll.cpp
Now I dont get any warnings but the compiler is stuck there, it wont give any result now

[cup]

Change

120: static simple_ptr<T> s_pObject;
121: static char s_objectState = 0;

to

120: extern simple_ptr<T> s_pObject;
121: extern char s_objectState;

Then look for misc.cpp and add

char s_objectState = 0;

[siddus]

Hi, this issue was solved after I removed static however, I have a new issue that seems to be AIX specific :

Just to summarize - I am using a cryptographic library called libcryptopp.a and building a wrapper around that library to access specific functions I need to use. The wrapper is in the form of a Shared Object so that I can call it from my legacy application.

I have libcryptopp.a and the functions in my wrapper (C++ program - aeswrap.cpp) are defined in the ".a" library. I am using the attached make file to build the Shared Object (libaes.so). However, when I try to call the function in my wrapper, I get a segmentation violation. It is worthwhile to mention that when I create an exe out of the same wrapper by statically linking it with libcryptopp.a it works fine.

I did a Truss on the file and these are the last few lines I got -

0509-151 The program does not have an entry point or
the o_snentry field in the auxiliary header is invalid.
kwrite(2, "\t 0 5 0 9 - 1 5 1 T h".., 115) = 115
_getpid() = 467000
lseek(3, 0, 1) = 4096
lseek(3, 4551, 0) = 4551
kread(3, "\t 0 5 0 9 - 1 9 4 E x".., 4096) = 454
0509-194 Examine file headers with the 'dump -ohv' command.
kwrite(2, "\t 0 5 0 9 - 1 9 4 E x".., 61) = 61
kfcntl(1, F_GETFL, 0x00000000) = 2
kfcntl(2, F_GETFL, 0x00000000) = 2
_exit(255)

ldd libaes.so gives -

/usr/lib/libcrypt.a(shr.o)
/usr/lib/libc.a(shr.o)
./libaes.so

and a dump gives -

libaes.so:

***Object Module Header***

1. Sections Symbol Ptr # Symbols Opt Hdr Len Flags

5 0x00080544 11231 72 0x3002
Flags=( EXEC DYNLOAD SHROBJ )
Timestamp = "25 Jun 18:23:48 2008"
Magic = 0x1df (32-bit XCOFF)

***Optional Header***
Tsize Dsize Bsize Tstart Dstart
0x0000c1c6 0x00001560 0x00000001 0x00000000 0x00000000

SNloader SNentry SNtext SNtoc SNdata
0x0004 0x0000 0x0001 0x0002 0x0002

TXTalign DATAalign TOC vstamp entry
0x0005 0x0003 0x00001308 0x0001 0xffffffff

maxSTACK maxDATA SNbss magic modtype
0x00000000 0x00000000 0x0003 0x010b RE

***Section Header Information***
Section Header for .text
PHYaddr VTRaddr SCTsiz RAWptr RELptr
0x00000000 0x00000000 0x0000c1c6 0x00000160 0x00077f5c

LN#ptr #RELent #LINent Flags
0x0007d20e 0x03d0 0x0889 0x00000020

Section Header for .data
PHYaddr VTRaddr SCTsiz RAWptr RELptr
0x00000000 0x00000000 0x00001560 0x0000c328 0x0007a57c

LN#ptr #RELent #LINent Flags
0x00000000 0x0475 0x0000 0x00000040

Section Header for .bss
PHYaddr VTRaddr SCTsiz RAWptr RELptr
0x00001560 0x00001560 0x00000001 0x00000000 0x00000000

LN#ptr #RELent #LINent Flags
0x00000000 0x0000 0x0000 0x00000080

Section Header for .loader
PHYaddr VTRaddr SCTsiz RAWptr RELptr
0x00000000 0x00000000 0x00005902 0x0000d888 0x00000000

LN#ptr #RELent #LINent Flags
0x00000000 0x0000 0x0000 0x00001000

Section Header for .debug
PHYaddr VTRaddr SCTsiz RAWptr RELptr
0x00000000 0x00000000 0x00064dd1 0x0001318a 0x00000000

LN#ptr #RELent #LINent Flags
0x00000000 0x0000 0x0000 0x00002000

I am stuck here and I dont know what I am doing wrong. Is the SO not being built correctly? I haave no clue what is to be done next. Just to debug I opened a file (FILE *fp) and noticed that the program exits after it encounters a particular function in the cpp program -

StringSource( PlainText.c_str() , true,
new StreamTransformationFilter( Encryptor,
new StringSink( CipherText )
) // StreamTransformationFilter
); // StringSource

This function is defined in the archive ".a" file.

I am just confused because it goes through when I make it into an exe? obviously there seems to be something wrong with the SO. Any help will be greatly appreciated.

AIX Release 1 Version 5
g++ version 3.3.2

[cup]

Just a long shot. Compile as position independent code and use gcc or g++ to do the linking

Changes

CFLAGS=-g -Wall $(INCLUDE) -fPIC
LDFLAGS=-shared
LD=gcc

It is something I've noticed about C++ progs - we can never get all the libraries whenever there is something STL related. If you use g++ or gcc or linking, the executable always works correctly. If you use ld, it sometimes works, but most of the time it comes up with a segv. If you run ldd on the exe, you'll find a whole load of stuff missing even though it links ok.

Be careful of -b in ld. Sun, AIX and gcc all have a -b parameter but the meaning and how it operates is different on all three.