GDB Use

Commonly used gdb Debugging command and usage

Be careful

When debugging the program, you need to add -g Options , It is also recommended to turn off the optimization options for compiler programs , Compiler program optimization generally includes 5 A level , namely O0-O4, among O0 It means no optimization , from O1-O4, The optimization level is getting higher and higher . Recommended O0.

start-up gdb Debugging method

Use gdb There are generally three ways to debug a program ：

1. Directly debug the target program ： gdb + filename

2. attach To the process ：gdb + attach + pid

   • When a program has started , But I want to debug the program , You need to gdb attch + Process to be debugged id. In the use of attach After reaching the target process , The debugger will stop , Use... Directly at this time continue Command the program to continue running .
   • If you want to end this debugging after debugging the program , And will not have any impact on the current process , Can be in gdb Command interface input detach Separation of command program and debugger .

3. debugging core file – Locate the process crash problem ：gdb + filename corename

   • linux By default, the system will not start when the program runs out core File function , have access to ulimit -c To check whether the system turns on this function . If you find that core file size yes 0 Words , Indicates that the build is closed core File options , Can pass ulimit + Option name + Set the value To modify the . It can be changed to the maximum allowable size , Such as 1024, Indicates the generation of the maximum core File size is 1024KB; It can also be changed to infinity , Such as ulimit -c unlimited

   • Be careful ： When the session is closed , This value will also become 0. If you want this option to be permanent , There are two ways to set ：

     1. stay /etc/security/limits.conf Add a line to the list ：

        #<domain> <type> <item> <value>
        *            soft     core     unlimited
        

     2. hold ulimit -c unlimited Add to /etc/profile In file , Just put it on the last line . Execute after successful modification source /etc/profile You can make the configuration take effect immediately . Of course, this only works for root user , If you want to use only one user , You can put ulimit -c unlimited Add to the corresponding of the user ~/.bashrc perhaps ~/.bash_profile In file

     3. Generated core The default file naming method is core.pid. The location is in the directory where the crash program is located .

   • If , When multiple programs crash at the same time , from core The name of the file corresponds to which service , How to solve it ？

     1. Mode one ： When the program starts, it is recorded pid

     2. Mode two ： Customize core Name and directory of the file

        • /proc/sys/kernel/core_uses_pid Can be controlled in the generated core Add to the file name pid As an extension , If added, the file content is 1, Otherwise 0

        • /proc/sys/kernel/core_pattern You can set the formatted core File save location or file name . The modification method is as follows ：

          echo "/corefile/core-%e-%p-%t" > /proc/sys/kernel/core_pattern
          

          Parameter name	Parameter meaning
          %p	add to pid To core In the filename
          %u	Add current uid To core In the filename
          %g	Add current gid To core In the filename
          %s	Adding results in core The signal to core In file
          %t	add to core File generation time （UNIX） To core In file
          %h	Add host name to core In the filename
          %e	Add program name to core In the filename

gdb Debugging multithreaded methods

1. Use gdb After running the program , Use ctrl + c Interrupt the program , Use info threads Command to see how many threads there are in the current process .

2. have access to thread + Switch the thread number to the corresponding thread , Use bt View the call stack of the thread

3. Sometimes in the case of multithreading , You only want to debug a thread , Instead of switching to another thread , You can use the options to lock the top note of the debug thread. scheduler-locking This option has three values , Namely on（ lock ）、step（ lock ）、off（ Release lock ）. Use as follows

   set scheduler-locking on/step/off
   

   • on and step difference ： Set to on when , Current thread execution next、step、until、finish、return On command , Other threads will not execute . and step If and only if next、step The command locks the current thread , Use until、finish、return Wait for debugging commands in the thread ( Non single step debugging command ), Other threads have a chance to run .

gdb Debug multiprocess methods

1. Method 1 ： Now? shell Use... In the window gdb Debug parent process , When a child process is fork After coming out , Opening a new one shell Window use gdbattach The order will gdb attach To child processes .

2. Mode two ：gdb The debugger provides a follow-fork Options , adopt set follow-fork mode Set up a process fork Create new subprocesses to make ,gdb Continue debugging parent or child processes .

   1.  Debug parent process ：set follow-fork-mode parent( default )
   2.  Debug child ：set follow-fork-mode child
   3.  Set debug mode ：set detach-on-fork [on|off]
      1.  The default is on; Indicates when debugging the current process , Other processes continue to run 
      2. off： When debugging the current process , Other processes are gdb Hang up 
   4.  View the process of debugging ：info inferiors
   5.  Toggle the current debugging process ：inferior +  process ID
   6.  Be careful ： To switch process debugging, you need to set the debugging mode to off
   

In this paper, the reference 《c++ The essence of server development 》

If this article helps you , Remember to connect three times with one button , One button three times laughing, ha ha , Excellent code capability ！

My ability is limited , If there is a mistake , Don't hesitate to correct ; Originality is not easy. , Welcome to reprint , Reprint please indicate the source !