如何让gdb重复获取stacktrace?

时间:2010-10-22 17:40:07

标签: batch-file profiling gdb stack-trace

喜欢

for((;;)) { 
    gdb -batch -n -ex 'set pagination off' -ex 'thread apply all bt' ffplay_g `pidof ffplay_g` >> /tmp/qq;
}

,但更快,没有每次重新加载GDB和符号?

Backtraces需要由计时器进行,而不是通过触发一些断点。

3 个答案:

答案 0 :(得分:2)

使用系统睡眠,这个gdb命令可以解决这个问题:

shell sleep 1

答案 1 :(得分:1)

根据评论If you want to stick with gdb, then why not script a gdb session? Your controller process can sleep for 50 ms, then wake up, send a ^C, t a a bt, c, and then go back to sleep. – Jeremy W. Sherman

的建议

http://vi-server.org/vi/bin/gdbdriver.pl

#!/usr/bin/perl -w

use strict;
use IPC::Open2;

my $init = "run";
my $command = "bt";
my $delay = 1;

my $need_int=0;

$init = shift @ARGV;
$delay = shift @ARGV;
$command = shift @ARGV;

die("Usage: gdbpriver.pl '' 0.1 'bt' gdb -q /path/to/proc 33344\n\tgdbdriver.pl init_command period_seconds backtrace_command startup_arguments\n") unless $ARGV[0];

my $pid = open2(\*OUT, \*IN, @ARGV);

print "pid=$pid\n";

print IN "set pagination off\n";
print IN "$init\n";

while(<OUT>) {
    if (/Starting program:/) {
    $need_int=1;
    last;
    }
    last if /\(gdb\)/;
}

sub intr() {
    kill 9, $pid;
    exit(0);
}
$SIG{'INT'} = \&intr;
sub spipe() {
    print "PIPE!\n";
}
$SIG{'PIPE'} = \&spipe;

if($need_int) {
    kill 2, $pid;
}

for(;;) {
    print IN "$command\n"; # backtrace
    print IN "c\n"; # continue the program
    while(<OUT>) {
    last if /Continuing./;
    print;
    }
    select undef, undef, undef, $delay; # sorry, nanosleep fails
    print "INT\n";
    kill 2, $pid; # SIGINT to gdb to make it interrupt the program
}

答案 2 :(得分:0)

附加,设置断点,在该断点上设置包含continue的命令,然后继续:

$ gdb attach ffplay_g
. . .
(gdb) b symbol
(gdb) comm 1
> t a a bt
> c
> end
(gdb) c

如果你问如何定期进行gdb中断,那么你可以循环调用步骤然后回溯,但是你不会很快得到很快:

(gdb) while 1
(gdb) t a a bt
(gdb) s
(gdb) end

如果您正在尝试分析您的流程,这是错误的方式。查看gprof或(在Mac OS / iOS下)Shark。