Use safe_mode when flag is set on boot for OS X #7
Conversation
this caused this test to usually fail.
| @@ -27,7 +27,7 @@ static bool pop(const void **buf, const void *end, | |||
| *buf += sizeof(*op); | |||
| *opc = op->opc; | |||
| if ((size_t) (end - *buf) <= op->namelen || | |||
| ((const char *) buf)[op->namelen] != '\0') | |||
| ((const char *) *buf)[op->namelen] != '\0') | |||
| return false; | |||
There was a problem hiding this comment.
without dereferencing buf, you would be checking if the namelen bytes after the pointer was equal to zero rather than the data.
which should never happen
| @@ -259,6 +273,8 @@ static int hook_posix_spawn_generic(__typeof__(posix_spawn) *old, | |||
| if (newp != newp_orig) | |||
| *newp++ = ':'; | |||
| newp = stpcpy(newp, dylib_to_add); | |||
| } else { | |||
| newp = stpcpy(newp, "\0"); | |||
There was a problem hiding this comment.
without the terminal operator, adjacent data would bleed into our string due to the memcpy(newp, p, next - p) where next, in the case of the last library would be p + strlen(p) which excludes '\0'
|
(wondering whether he was especially tired when writing this code or whether all his code is this bad) The uninitialized bool is correct, though. |
|
Right, thanks for looking at my code! Now that I'm actually working on this - I added the critical bug fixes from this PR in slightly modified form, which I will push once I check that they actually work. The Safe Mode thing makes sense, and I'll merge it if you fix it up as I requested. |
|
done (and squashed) |
we want to compare against p, not next
|
I guess I might as well merge the commits as is since it's not like my commit history isn't total shit, but thanks to Dolphin I'm used to being a stickler on PRs, and that is not rebased properly. :p |
When the user boots into safe mode with the -x flag or otherwise on OS X, substitute should be disabled. Also, there's a commit in here which makes the bootstrap stuff build no matter what (mostly so I could test on OS X, which I'm still in the process of setting up a VM to do so), I don't know if that goes according to your plan for this project, though.
I haven't tested this bit of code in the context of substitute, but I can affirm that it otherwise works as far as detecting safe mode.
Also, this has a bug fix for bundle-loader.c in which the test for the char after namelen characters in the name is equal to '\0' would usually fail.