<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#ffffff">
On 31.05.2011 19:45, Devang Patel wrote:
<blockquote
cite="mid:28E8AD70-77FD-4FB3-A37C-E3A30284047D@apple.com"
type="cite"><br>
<div>
<div>On May 31, 2011, at 10:36 AM, <a moz-do-not-send="true"
href="mailto:trash-stuff@gmx.de">trash-stuff@gmx.de</a>
wrote:</div>
<br class="Apple-interchange-newline">
<blockquote type="cite"><span class="Apple-style-span"
style="border-collapse: separate; font-family: Verdana;
font-style: normal; font-variant: normal; font-weight:
normal; letter-spacing: normal; line-height: normal;
orphans: 2; text-indent: 0px; text-transform: none;
white-space: normal; widows: 2; word-spacing: 0px;
font-size: medium;">On 31.05.2011 19:22, Devang Patel wrote:
<blockquote
cite="mid:63165691-E116-4435-9188-7976D35830BB@apple.com"
type="cite"><br>
<div>
<div>On May 30, 2011, at 11:11 AM,<span
class="Apple-converted-space"> </span><a
moz-do-not-send="true"
href="mailto:trash-stuff@gmx.de">trash-stuff@gmx.de</a><span
class="Apple-converted-space"> </span>wrote:</div>
<br class="Apple-interchange-newline">
<blockquote type="cite"><span class="Apple-style-span"
style="border-collapse: separate; font-family:
Verdana; font-style: normal; font-variant: normal;
font-weight: normal; letter-spacing: normal;
line-height: normal; orphans: 2; text-indent: 0px;
text-transform: none; white-space: normal; widows:
2; word-spacing: 0px; font-size: medium;">Hi all,<br>
<br>
I am processing DWARF line and column information in
(x86 and ARM) executables in order to produce a
mapping from the machine instructions back to the
original source code (C/C++). Using the line numbers
is quite straightforward ("libdwarf" [1] is doing
the work me.) But when comparing the column numbers
(extracted from the DWARF line table) with the
corresponding source code locations, it becomes
clear that they are not very "useful".<br>
<br>
Consider the following small example (C++):<br>
<blockquote><tt> 1: #include <iostream><br>
2: #include <ctime><br>
3: #include <cstdlib><br>
4: using namespace std;<br>
5: int main() {<br>
6: int j = 0; cin >> j; long sum = (j
< 0 ? -5 : 4) + rand();<br>
7: for(int i = 0; i < j; i++) { sum +=
j*j-2; cout << (sum / 2) << endl; }<br>
8: srand(time(NULL));<br>
9: double d = rand() / 10.341; int t =
(int)d+j*sum;<br>
10: cout << sum << d << t
<< j;<br>
11: return (0);<br>
12: }</tt><br>
</blockquote>
Compiling this with "clang++ Main.cpp -g -O3 -o
column" result in the following location information
within the generated executable:<br>
<blockquote><tt>$ dwarfdump -l column<br>
<br>
.debug_line: line number info for a single cu<br>
Source lines (from CU-DIE at .debug_info offset
11):<br>
<source file> [line,column]
<pc> //<new stmt or basic block<br>
.../locale_facets.h: [868, 2] 0x80488f0 //
new statement</tt><br>
<tt> [...]</tt><br>
<tt>.../Main.cpp: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>[
8, 2] 0x804896f // new statement</tt><br>
<tt>.../Main.cpp: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>[
9,28] 0x8048983 // new statement</tt><br>
<tt>.../ostream: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>
[165, 9] 0x8048990 // new statement</tt><br>
<tt>.../Main.cpp: </tt><tt> </tt><tt><span
class="Apple-converted-space"> </span>[
9,28] 0x80489a0 // new statement</tt><br>
<tt>.../ostream: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>[209,
9] 0x80489ac // new statement</tt><br>
<tt>.../Main.cpp: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>[
9,28] 0x80489b5 // new statement</tt><br>
<tt>.../ostream: </tt><tt> <span
class="Apple-converted-space"> </span></tt><tt>[209,
9] 0x80489bb // new statement</tt><br>
<tt> [...]</tt><br>
<tt>.../basic_ios.h: [ 48, 2] 0x8048a23
// new statement // end of text sequence</tt><br>
</blockquote>
Now, have a look at source code line 9. The
extracted debug info above says that we've 3
"instruction sets" (beginning at<span
class="Apple-converted-space"> </span><tt>0x8048983,<span
class="Apple-converted-space"> </span></tt><tt>0x80489a0</tt><span
class="Apple-converted-space"> </span>and<span
class="Apple-converted-space"> </span><tt>0x80489b5</tt><span
class="Apple-converted-space"> </span>respectively)
which correspond to line 9. But all of them are
labeled with column number 28! According to my
understanding, this does not contribute any further
information to support my task (= mapping assembler
code back to the source lines or even to statements
within a line). Did i miss anything?<br>
</span></blockquote>
</div>
<br>
<div>You are looking at the line table produced at -O3,
i.e. after aggressive optimizer had opportunities to
optimize code. Try -O0 and see if it helps.</div>
</blockquote>
First of all, thanks for your reply!<br>
<br>
I've already checked that at -O0 but it results in the same
information.</span></blockquote>
<div><br>
</div>
<div>You mean, the instructions with given line and column
number do not match the source code construct at that location
? <br>
</div>
</div>
</blockquote>
No, they do.<br>
<blockquote
cite="mid:28E8AD70-77FD-4FB3-A37C-E3A30284047D@apple.com"
type="cite">
<div><br>
<blockquote type="cite"><span class="Apple-style-span"
style="border-collapse: separate; font-family: Verdana;
font-style: normal; font-variant: normal; font-weight:
normal; letter-spacing: normal; line-height: normal;
orphans: 2; text-indent: 0px; text-transform: none;
white-space: normal; widows: 2; word-spacing: 0px;
font-size: medium;"> (The documentation about "Source Level
Debugging with LLVM" says "<b>LLVM debug information always
provides information to accurately read the source-level
state of the program, regardless of which LLVM
optimizations have been run</b>, and without any
modification to the optimizations themselves." [1])<br>
</span></blockquote>
<br>
</div>
<div>It means the instructions with given line and column number
matches the source code construct at that line/col number. It
does not mean that optimizer/code generator will not reorder
instruction. It also does not mean that optimizer/code generator
will not emit instruction without line number information. It
means, if there is a line number information, it is as accurate
as possible to map source construct.</div>
</blockquote>
Yes, that matches my understanding, too. But I thought that clang
would be able to emit <b>more</b> than one (different) column
number per line. As in my example, for line number 9 (in Main.cpp),
there are <b>three</b> entries in the DWARF line table. But all of
them contain the <b>same</b> information. As a consequence, the
associated assembler instructions were all mapped to the same source
line and thus, the column information is useless...? I mean, what
are the additional information included in the column numbers?<br>
<br>
I extracted the assembler instructions for the 9th line (x86):<br>
<tt>.../Main.cpp: 9<br>
double d = rand() / 10.341; int t = (int)d+j*sum;<br>
^<br>
8048983: e8 40 fe ff ff call 80487c8
<rand@plt><br>
8048988: 89 c7 mov %eax,%edi<br>
804898a: 8b 5d f0 mov -0x10(%ebp),%ebx<br>
804898d: 0f af de imul %esi,%ebx<br>
80489a0: f2 0f 2a c7 cvtsi2sd %edi,%xmm0<br>
80489a4: f2 0f 5e 05 f0 8a 04 divsd 0x8048af0,%xmm0<br>
80489ab: 08 <br>
80489b5: f2 0f 2c f0 cvttsd2si %xmm0,%esi<br>
80489b9: 01 de add %ebx,%esi</tt><br>
<br>
I hope that makes it clearer... ;-)<br>
<br>
BTW, any hints to my cross-compilation-related question?<br>
<br>
Best regards<br>
Adrian<br>
</body>
</html>