stl_test not working properly on Linux? #534
Comments
|
From [email protected] on August 10, 2011 07:33:15 I'm not very familiar with DR on Linux - am I right that running libXXX.so.debug is wrong?
|
|
From [email protected] on August 10, 2011 07:36:04 I believe .debug is the file storing the debug symbols. |
|
From [email protected] on August 10, 2011 07:37:43 OK then why doesn't it call me crazy right away? :) |
|
From [email protected] on August 10, 2011 07:44:31 Same result on this simple test: void event_exit() {} DR_EXPORT void dr_init(client_id_t id) { |
|
From [email protected] on August 10, 2011 07:45:53 FTR, this is all happening on r925 |
|
From [email protected] on August 10, 2011 07:53:43 oops, looks like "-DCMAKE_BUILD_TYPE=Debug" is no-op for DR. However, I get the same segfault when building with -DDEBUG=ON |
|
From [email protected] on August 10, 2011 07:57:32 .debug file itself is also a valid shared library file with much more debug information. |
|
From [email protected] on August 10, 2011 08:11:33 libstl_test.so passes when I remove all cout/cin/iostream lines from stl_test.cpp |
|
From [email protected] on August 10, 2011 08:21:14 for future linux bug reports, what we need is info on the crash, and whether it works w/o private loader this does work with -no_private_loader the crash involves private loader TLS: (gdb) bt (gdb) disas __btowc (gdb) info reg Status: Accepted |
|
From [email protected] on August 25, 2011 13:00:35 Status: Fixed |
|
From [email protected] on August 26, 2011 03:02:09 Better now ( r951 ) but still seg.faults on exit: $ ./bin32/drrun -client ./api/samples/bin/libstl_test.so 0 "" ~/sandbox/hello The emptycpp.cpp client is OK now Status: Started |
|
From [email protected] on August 26, 2011 08:25:36 Is it possible to provide more information about either the seg fault context, or how I can reproduce the problem? |
|
From [email protected] on August 26, 2011 08:27:45
|
|
From [email protected] on August 26, 2011 08:33:26 ../exports/bin32/drrun -debug -ops "-loglevel 0 -msgbox_mask 0x0" -client ../exports/samples/bin32/libstl_test.so 0 "" ../../test/hello-32b It worked on my machine. It looks the segfault happens before the last Exit is printed, so something is wrong while libraries calling their fini functions |
|
From [email protected] on August 26, 2011 08:45:10 Hm, looks like I have a release build (?!) |
|
From [email protected] on August 26, 2011 08:47:24 Oh, right: -DDEBUG=ON, not -DCMAKE_BUILD_TYPE=Debug Can you try a release build? |
|
From [email protected] on August 26, 2011 09:04:14 Yes, the release build cause the problem. I will look into it. |
|
From [email protected] on September 28, 2011 21:15:19 Can you check if it works in the TOT revision. |
|
From [email protected] on September 30, 2011 04:43:19 I think it works for me now. |
|
From [email protected] on September 30, 2011 08:19:29 closed by r1003 , which fixes the issue i555, client system call using sysenter. Status: Fixed |
|
From [email protected] on September 30, 2011 08:46:28 Just curious: did I help you find a sysenter bug or is it just a coincidence? |
|
From [email protected] on September 30, 2011 08:53:08 It started with this issue, and I tried to reproduce the problem without using c++ library, and find the issue #555 . I am still not sure if this issue is completely resolved in all platforms. |
From [email protected] on August 10, 2011 10:32:50
This is Ubuntu Lucid x64
$ cat hello.c
#include <stdio.h>
int main(void) {
printf("Hello, world!\n");
return 0;
}
$ gcc -g -m32 hello.c -o hello
$ cd dynamorio
$ mkdir debug && cd debug
$ ld --version
GNU ld (GNU Binutils for Ubuntu) 2.20.1-system.20100303
$ CXXFLAGS=-m32 CFLAGS=-m32 cmake -DCMAKE_BUILD_TYPE=Debug .. && make -j10
...
$ ./bin32/drrun -client ./api/samples/bin/libbbcount.so 0 "" ~/sandbox/hello
Client bbcount is running
Hello
Instrumentation results:
1809 basic block executions
113 basic blocks needed flag saving
396 basic blocks did not
$ ./bin32/drrun -client ./api/samples/bin/libstl_test.so 0 "" ~/sandbox/hello
Segmentation fault <----- !!!!!!!!!
Original issue: http://code.google.com/p/dynamorio/issues/detail?id=534
The text was updated successfully, but these errors were encountered: