Support interposing a (very basic) gdb script that sets debug file paths in the sources command.
#3714
Conversation
We should do something about that, but not in this PR. |
| index = strcspn(buf, "\n"); | ||
| buf[index] = 0; | ||
|
|
||
| char* token = strtok(buf, delimiter); |
There was a problem hiding this comment.
Are we going to regret not being able to handle paths containing ':'?
I'd prefer an approach that doesn't barf on special characters in file names. Maybe \n is already a lost cause, but other than that...
There was a problem hiding this comment.
Using the path separator here comes from gdb, not from me. Using colons in directory names in Linux is already heavily frowned upon because you can't e.g. add them to PATH. I don't think we'll regret this.
src/SourcesCommand.cc
Outdated
| FATAL() << "Failed to wait on gdb script host"; | ||
| } | ||
|
|
||
| if (!(WIFSTOPPED(status) && WSTOPSIG(status) == SIGSTOP)) { |
There was a problem hiding this comment.
This is weird! why can't we just communicate over the pipe like a regular program?
There was a problem hiding this comment.
I'm a bit afraid this will interact poorly with ctrl-Z job control
There was a problem hiding this comment.
Mmm yeah. I was in the mindset of writing kernel behavior tests :P
There was a problem hiding this comment.
Are you going to change this? I don't like it :-)
src/SourcesCommand.cc
Outdated
| FATAL() << "Failed to wait on gdb script host"; | ||
| } | ||
|
|
||
| if (!(WIFSTOPPED(status) && WSTOPSIG(status) == SIGSTOP)) { |
There was a problem hiding this comment.
Are you going to change this? I don't like it :-)
…ths in the `sources` command.
khuey
force-pushed
the
gdb-symbol-loading-script
branch
from
6ea338f to
a0ecb21
Compare
This is relatively straightforward, just ugly. One notable difference in the behavior from gdb is that when multiple paths are provided gdb will try them all in sequence but this code just tries the last one. That works for the use case that's motivating this PR but may cause issues down the line.
I'm reasonably confident I didn't regress anything but there's no automated testing for
rr sources.