new fork() and clone() approach

used PTRACE_O_TRACE{FORK,VFORK,CLONE} to trace newly created
processes instead of figuring it out whether a given syscall
would create another process or not

1 files changed, 6 insertions, 8 deletions

diff --git a/TODO b/TODO
index 4048185..070a680 100644
--- a/TODO
+++ b/TODO
@@ -14,17 +14,11 @@
 * More architectures, cleaner way to port
 * More operating systems (solaris?)
 * Option -I (inter-library calls)
-* Modify ARGTYPE_STRING[0-5] types so that they not stop displaying chars when '\0' is encountered
+* Modify ARGTYPE_STRING[0-5] types so that they don't stop displaying chars when '\0' is seen
 * Get rid of EVENT_ARCH_SYSCALL and EVENT_ARCH_SYSRET
-* EVENT_FORK, EVENT_CLONE, EVENT_EXEC instead of fork_p() and exec_p()
-* If EVENT_FORK is received:
-  + Add the new process to the list, with a state of "future child of XXX"
-* If EVENT_CLONE is received:
-  + Add the new process to the list, with a state of "future clone of XXX"
+* EVENT_EXEC instead of exec_p()
 * If EVENT_EXEC is received:
   + Clean structs with breakpoints, open new created program
-* If a signal is received from an an unknown process, add it to the list,
-  with a state of "new process"
 * Cleaner way to use breakpoints:
   + BP is placed in the PLT
   + When control hits there:
@@ -42,3 +36,7 @@
     without having to remove it
 * Get rid of GNU's autoconf stuff
 * List source dependencies in Makefile
+* Create different ltrace processes to trace different children
+* debug: change "-d" option to be something like "-d elf,events", or "-d breakpoints"
+* Find out if a process is sharing memory with its parent?
+* When using -p, find out if that process is sharing memory with other processes