LLVM IR source line error

Project Infrastructure LLVM Dev List Archives
#1

Hi

I am using LLVM to help me do some code analysis. I wrote a LLVM Function Pass to help me to generate some information. I use the code below to get the source line information for every instruction.

for (BasicBlock &BB : F){

for(Instruction &I: BB){
DILocation* Loc = I.getDebugLoc().get();

unsigned Line = Loc->getLine()

}
}

I think the function’s first block’s first instruction’s source line should be the beginning of the function. Most of the functions inside the binary follow the rules. However, I found the file system.h in the coreutils-8.28 has a function like below:

611 static inline void
612 emit_backup_suffix_note (void)
613 {
614 fputs (("
615 \n
616 The backup suffix is ‘~’, unless set with --suffix or SIMPLE_BACKUP_SUFFIX.\n
617 The version control method may be selected via the --backup option or through\n
618 the VERSION_CONTROL environment variable. Here are the values:\n
619 \n
620 "), stdout);
621 fputs (("
622 none, off never make backups (even if --backup is given)\n
623 numbered, t make numbered backups\n
624 existing, nil numbered if numbered backups exist, simple otherwise\n
625 simple, never always make simple backups\n
626 "), stdout);
627 }

When analysis this function. The first block’s first instruction’s source line is 614 rather than 612 or 613. Is it a bug or any other meaning resulting in such case. Below is the IR. The first IR’s source line if 614.

%call = call i8* @gettext(i8* getelementptr inbounds ([221 x i8], [221 x i8]* @.str.42, i32 0 , i32 0)) #12, !dbg !1295
%0 = load %struct._IO_FILE*, %struct._IO_FILE** @stdout, align 4, !dbg !1295
%call1 = call i32 @fputs_unlocked(i8* %call, %struct._IO_FILE* %0), !dbg !1295
%call2 = call i8* @gettext(i8* getelementptr inbounds ([222 x i8], [222 x i8]* @.str.43, i32 0, i32 0)) #12, !dbg !1296
%1 = load %struct._IO_FILE*, %struct._IO_FILE** @stdout, align 4, !dbg !1296
%call3 = call i32 @fputs_unlocked(i8* %call2, %struct._IO_FILE* %1), !dbg !1296
ret void, !dbg !1297

Can anyone help me or give me some suggestions? Many Thanks

Regards
Muhui

#2

Hi

I am using LLVM to help me do some code analysis. I wrote a LLVM Function Pass to help me to generate some information. I use the code below to get the source line information for every instruction.

for (BasicBlock &BB : F){

for(Instruction &I: BB){
DILocation* Loc = I.getDebugLoc().get();

unsigned Line = Loc->getLine()

}
}

I think the function’s first block’s first instruction’s source line should be the beginning of the function.

I’m not sure why that would necessarily be true - if there are no instructions associated with the beginning of the function. (no frame setup, no allocas, etc)

Most of the functions inside the binary follow the rules. However, I found the file system.h in the coreutils-8.28 has a function like below:

611 static inline void
612 emit_backup_suffix_note (void)
613 {
614 fputs (("
615 \n
616 The backup suffix is ‘~’, unless set with --suffix or SIMPLE_BACKUP_SUFFIX.\n
617 The version control method may be selected via the --backup option or through\n
618 the VERSION_CONTROL environment variable. Here are the values:\n
619 \n
620 "), stdout);
621 fputs (("
622 none, off never make backups (even if --backup is given)\n
623 numbered, t make numbered backups\n
624 existing, nil numbered if numbered backups exist, simple otherwise\n
625 simple, never always make simple backups\n
626 "), stdout);
627 }

When analysis this function. The first block’s first instruction’s source line is 614 rather than 612 or 613.

The first instruction is the call to gettext (presumably expressed by the _ macro) that is on line 614. Doesn’t look like there’s any instruction that could/should be attributed to 612 or 613.

The DISubprogram describing the function should probably have some a “scopeLine” attribute of 613, which might be what you want.

#3

Hi David

Thank you very much for your suggestions. ScopeLine works well. It could represent the start of a function. Do you know whether I can get the start column of this function? Many Thanks

Regards
Muhui

#4

I don’t believe so, no. The ScopeLine was added for adjusting profile samples & was just what was needed for that task - I don’t think there’s any equivalent column info.