| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
|
| |
|
|
| |
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The smallest data element is:
struct crash_item { char *content; unsigned flags; };
where content is, eh, content, and flags is a bit flag field.
crash_data_t is a map of crash_item's, implemented as a pointer
to heap-allocated GHashTable.
vector_of_crash_data_t is a vector of crash_data_t's, implemented
as a pointer to heap-allocated GPtrArray.
Most operations have light wrappers around them to hide the nature
of the containers. For example, to free vector_of_crash_data,
you need to use free_vector_of_crash_data(ptr) instead of
open-coding g_ptr_array_free. The wrapper is thin.
The goal is not so much to hide the implementation, but more
to make it easier to use the correct function.
dbus (un)marshalling functions convert crash_item to three-element
array of strings, in order to keep compatibility with abrt-gui (python).
This can be changed later to use native representation.
crash_data_t and vector_of_crash_data_t are represented in
"natural" way, no funny stuff there.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch minimally affects code per se - it adds
create_crash_dump_dir() function which takes in-memory
representation of the dump (map_crash_data_t object)
and creates an on-disk representation.
Then it returns pointer to struct dump_dir.
With this function, it is possible to run an event
on a user-constructed map_crash_data_t:
map_crash_data_t cd;
add_to_crash_data(cd, "foo", "bar");
struct dump_dir *dd = create_crash_dump_dir(cd);
char *dir_name = strdup(dd->dd_dir);
dd_close(dd);
run_event(run_state, dir_name, event);
delete_crash_dump_dir(dir_name);
which is, basically, what report library is about.
The biggest part of the patch is reshuffling of
header files, with the following result: three header
files which are not cluttered by other ABRT stuff,
and expose the following API each:
crash_dump.h - in-memory crash dump data structs and ops
dump_dir.h - on-disk crash dump data structs and ops
run_event.h - run_event() and friends
These is a test application, test_report.cpp.cpp,
which demonstrates (and tests) usage of these headers.
(grep for "test-report" and enable it in build system
if you want it to be built).
It creates a temporary dump in
/var/run/abrt/tmp-NNNN-NNNNN and runs "report" event.
Deleting of temp dump is disabled for testing purposes -
you might want to see the contents.
Here is how the binary looks like:
text data bss dec hex filename
3730 668 48 4446 115e /usr/bin/test-report
linux-vdso.so.1 => (0x00007ffffa80f000)
libabrt.so.0 => /usr/lib64/libabrt.so.0 (0x00007fad1f35e000)
libglib-2.0.so.0 => /lib64/libglib-2.0.so.0 (0x0000003c1f200000)
libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x0000003c2c200000)
libm.so.6 => /lib64/libm.so.6 (0x00007fad1f0da000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x0000003c28e00000)
libc.so.6 => /lib64/libc.so.6 (0x00007fad1ed5b000)
/lib64/ld-linux-x86-64.so.2 (0x00007fad1f570000)
Next step would be to clean up the namespace,
then rename libabrt.so into libreport.so,
then split it into a separate package so that we can install it
without installing other two abrt .so files.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
|
|
|
Some files and functions are renamed, no logic changes.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
|
