Programming :: Get A Core Dump In One Of The C Library Calls (like Strcpy Or Printf)
Sep 20, 2010
I am trying to port some "C" code from Solaris to Linux. I have a Dell PowerEdge R610 with an Intel Xeon E5504 quad core processor running Red Hat Linux Enterprise 5.3. I am compiling in 64 bit mode. I have managed to get the code compiled and linked, but when I attempt to execute it, I get a core dump in one of the C library calls (like strcpy or printf.)
I have a static library that contains our own code that makes the call to the C library. If I move the library method into the source file with the main method and rename it to be certain that I am executing my method instead of the method in our library, the call succeeds. Eventually another static library call is made that results in a core dump in the shared object. I compile my library code into a static library with gcc as:
I want to generate core dump files from my program when it crashes. Its a pretty big process and has about 10-11 threads in it.I have followed the documentation to enable core dump by setting ulimit to unlimited etc. I quickly tried "A demo program creating a core dump" from the following webpage, which succeeds in Segfault and dumping a core file in the directory that I configured.However, I tried running my original program and caused it to crash. I did this by making calls to kill(), raise() or the same null pointer access as shown in the webpage above. In each case, my program crashed but did not generate a core dump file. Am I missing something?My program is in C++ and my environment is Redhat 9.0 (kernel 2.4.20)
Going through the "Why do I NOT get a core dump?" section on the same webpage as above, I can see two potential problems. One - there are issues with the suid/sgid (bullet # 6). I am not able to change any settings with suid because my system does not contain either /proc/sys/fs/suid_dumpable or /proc/sys/kernel/suid_dumpableTwo, my program has threads in it and the bullet # 8 is the problem.
I'm trying to write a C program that extends an array to any user inputed size.
Code: if (arraysize == 0) { arraysize = (int) pos + 1; a = (int *) calloc (arraysize,sizeof(int)); for (i = 0 ; i < arraysize ; i++ ) a[i] = -1; code....
The program dumps with that sequence of inputs everytime, but might dump an input before or after if different positions are requested. Interestingly, when I tested pos = 2000..2008, I got no dumps. So is realloc somehow trying to extend the array into bad space?
I am developing an application whose executable is generated inside a certain folder hierarchy (say: /DevPath/MyProject/bin). My source code is located in a different branch of this hierarchy (say: /DevPath/MyProject/src). When my app crashes, its core files are stored in /DevPath/MyProject. I'm developing the app on a pc, but running it on a separate platform in which i can only execute it. Folder hierarchy is the same as above on both computers. Usually, when a new executable version is ready, we update both the executable and the source code on the target platform, transferring all the new /DevPath/MyProject folder on it. But sometimes it can really be a bother, so we update only the executable.
1)In the case we only update the executable, keeping an old source code version, and the app generates a core file, can i trust the backtrace produced by gdb in this case? I.e., does gdb need the latest source code files or it just needs the debugging information?
2) (More radical question ) Do i really need to keep the source code on the target platform for core dump analysis or i just need the executable?
In one of our core dump we have the followings in the core back trace:
#0 0xb77bf947 in raise () from /lib/tls/libc.so.6 #1 0xb77c10c9 in abort () from /lib/tls/libc.so.6 #2 0xb77f56ba in __fsetlocking () from /lib/tls/libc.so.6 #3 0xb77fcf7f in mallopt () from /lib/tls/libc.so.6 #4 0xb77fd022 in free () from /lib/tls/libc.so.6
It occurred in a memory block free operation. From our analysis, there seems no issue relate the the memory block it self. The memory pointer pointed to the right memory block to be freed and the contents of the memory seems right (not corrupted), in one world, there is nothing obviously wrong. Does any one have any ideas what could be wrong when seeing about?
To analyse a coredump, I need to specify program name/path in GDB/KDevelop. Since the program name along with arguments is also within a core dump, I wonder if it doesn't keep the proper path of program that crashed and so asks for it?
Some context: I'm calling some functions from within a C program and want to measure how long they take to execute. I've done this successfully, calling printf to print the results to the screen. This is reasonably fast. Ideally I'd like these to be dumped to a file, but fprintf presumably has some overhead since it writes to disk? Unless the data is buffered by the kernel and flushed later? What kind of delay can I expect before fprintf returns?
If I redirect the stdout to a file when I instantiate the binary as a process, will calls to printf also experience any overhead from redirection to a file? My aim here is to capture all data in a file, with the minimal amount of coding and effort, but with minimal impact in terms of time overheads for printf calls.
We recently ported an application over from IRIX to Linux, specifically SUSE Enterprise 11 and gcc (glibc 2.9). On IRIX there was a library trace utility call 'par' that allowed me to execute a C program and trace all the library calls it made, tracking how much time was spent in each call. I have been trying to find something similar for SUSE. I have found the ?trace programs (strace, ltrace, and ktrace) none of which is default installed on SUSE 11.Does any one have a suggestion on which ?trace to install and use or is there something else out there better?
What is the real difference between system calls and normal function calls. Ultimately function calls too would be passed to kernel for some or the other work.
In my program, I fork() to get a child process. Because of some problem, child process terminates by a segmentation fault. Parent process is still running. I have compiled my code with -g option. I have done: ulimit -c unlimited. I am not getting core dump of the child process. How can I get the core dump of child process?
I am using RHEL 4.7 (32bit) on HP Proliant 380G6 series server. We are using Electric Cloud Agents on these servers. Nowadays we are facing some memory issues and its creates some kernel panic and then restarts the server. When i reported the issue to my application team, they asked me to come with the core dump. I googled it enough, then i set ulimit value as unlimited. (previously it was 0, then i made a entry in /etc/profile file as follows ulimit -c unlimited) But still whenever my server restarts due to that kernel panic, it couldnt generate the core dump. My application was installed on /opt
I have used Dump Command to dump the application files. For Full backup the level 0 is working fine. For incremental backup I used the level 1 or 2 it is getting the error as
DUMP: Only level 0 dumps are allowed on a subdirectory DUMP: The ENTIRE dump is aborted.
The code I used =============================== #!/bin/bash #Full Day Backup Script #application folders backup #test is the username now=$(date +"%d-%m-%Y") [Code]...
i just touch linux, may i know how can i convert the core dump file to a readable textfile, which include all the information, which is in core dump, such as all variables, threads information, call trace for each tasks, and so on. i know use the GDB can view this, but it won't dump all the informations to one text file. but sometimes, people want to view the core dump reason without Linux environment.
Id like to share with you a strange behavior I�ve been seen with the printf C language function running in Fedora.The code opens a file as showed:descString = fopen( strcat( stringName, txt), a+);There are also some printf functions in the code.However, instead of the printf functions to print their content at the screen, theyve been printing them in the file with the descString file descriptor.Have you ever seen something like that?To print something to the file I use the fprintf function. I didn't make a mistake to confuse printf with fprintf. I'm sure.
For eg: $NUMBEROFPASSES=6.19 to round up I use NUMBEROFPASSES=$(printf %.0f $NUMBEROFPASSES) What I need to do is round up from eg: 6.10 to be 7 and if lower than 6.10 round down to 6
I got curious during the construction of my program to search for blocks of text that contain certain words, and wanted to print out blocks of text with the words I was searching for, initially, in ALL CAPS and seperated by tabs for easy identification. That got boring, though, so I thought, "Can printf do colors, too?" Yes it can.... Who knew?
Assume someone bind a particular process to a particular CPU core(In multi core machine) by using sched_setaffinity() like functions. Then how we can get that process running core id and CPU core utilisation of that process on that running CPU core(Pragmatically or by a Linux command)?.
I'm reading about shared, static, and dynamic libraries. What is SDL? Is it static, shared, or dynamic?
I always thought a library would be a lot of .h and .cpp files compiled separately into .o files and then if you compiled your own program you could use the -l parameter to link the library and it was all compiled together. Now I'm not so sure.
I don't even see any SDL .cpp files in my system anywhere. All I have are lots of SDL .h files in /usr/include/SDL and I don't really understand the code in them.
I'm making a wild guess here: SDL is a shared library. SDL itself is NOT compiled into my program, therefore SDL must be on any system my program tries to run on. When I compile and link SDL all it needs is the header files to know what SDL function and objects it can use. And then on every system it uses an already compiled SDL shared library thingy somewhere.
So... where is that part of SDL? All I can find are header files.
I'm thinking the advantage of shared libraries is that someone could say update SDL on their own system and take advantage of the new features without having to download new executables with the new version of SDL compiled into them for every program that uses SDL.
So if I'm making an editor and a game engine and they both use a lot of the same .cpp and .h files that I wrote and I'm tired of updating one and then the other and I need to turn them into a library, then a shared library might be kind of a silly solution. I could just make a static library. Right? Because it's not SDL. Nobody else is ever going to use this library.
How can we convert a dynamic library (filename.so) to a static library (filename.a) using gnu gcc . Can we get a static library form a dynamic library . I saw a few post in which the conversion form a static library to a dynamic library is mentioned but, unfortunately, not the other way.
I have a process which runs for a while and then due to some unknown reason ends abruptly without giving a core. I tried looking in the logs and after watching for a pattern am still not 100% sure of the reason. So I was wondering if there is a way to catch the signal which ends the process and print some values in the handler function. I have called the handler for SIGTERM and SIGABRT functions but none of these are getting triggered. I looked online and did not find any other option. Can you please suggest if tehre is any other signal that can be caught for this unknown abrupt termination.
I have a problem in making multiple calls using PJSIP. I registered different accounts that implements IAccount and added that in the CallManager. When i try to make the calls from the same CallManager, I could able to make the first call. When second call is made the first call got cut and it is alerting and the same for the thrid call..Could anyone help me in this regard.
Note: Calls has to be made from different numbers to different destinations. Is it possible to make multiple outbound calls simultaneously from the same instance of CallManager?
I know there is a way to call winapi in runtime in windows. I want to ask how can I call a system call in gcc in runtime (when I don't know what it can be)? I don't mean syscall that I think is for calling only system calls and not library functions.
Is there a way to hook calls to new/malloc, delete/free in C++? I tried the following methods: LD_PRELOAD -> fails for malloc, because dlsym seems to depend on malloc -Wl,--wrap,malloc -> doesn't work for 'new' gcc hooks -> doesn't work always, e.g. for uclibc there are no hooks Are there any other methods I could try?
