blob: 89c1574c9137b0d22eec12f96a329520a1a0a264 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
|
o Shouldn't we separate command parsing from a service implementation?
- context.c
o event_pid_stopped() using the PTRACE_EVENT_CLONE to determine if the new context is a thread is not correct. The only
way I know of is to look at the Tgid field of /proc/<pid>/status
- runctrl.c
o command_resume() and command_step_into() share so much code that they could be unified.
o Sometimes you flush “broadcast_stream.flush(&broadcast_stream);”. What is the logic/rules?
o VxWorks: Need Component Description File for the agent: C:\WindRiver-2.6\vxworks-6.5\target\config\comps\vxWorks
11/1/2007
- It would be really good to put comments on at least the all global functions in headers.
- Complex structures would also be very good to put comments on. For example in breakpoints.c
- Somehow we should make it clear what needs to be done to add another transport layer.
Perhaps have a template or a readme file for it.
- breakpoints.c, line 190: if (!bi->ctx->exited && bi->ctx->stopped) {
o Why is bi->ctx->stopped test needed? It looks like we forget to unpatch the instruction is the target
happens to be running when we get to this point since after the if the “planted” flag is unconditionally
cleared. Shouldn’t this be an assert instead?
o Similar problem if context_write_mem() results in error then we clear the planted flag even though we haven’t
really unplanted it.
- streams.h:
o It would be good if the InputStream and OutputStream was designed so characters could be read/written without
causing a function call in the normal case. Something similar to stdio getc/putc.
- runctrl.c, line 326
o Should it not return the error code to the client?
- stacktrace.c
o line 176: Don’t we need to check code[0] in this case?
o Line 242: no error check on strtol. Why not used id2pid?
|
