You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Reconsider debugger discovery for debuginfo test suite #148361
E.g. don't discover some random lldb from PATH when it's not explicitly configured, maybe explicitly asking for which debugger versions the user wants is the way-to-go, even if that requires more initial setup.
One of the follow-ups after #148099 will be to move the gdb and cdb discovery (which do the same thing) out into bootstrap, so that at least all of the questionable debugger discovery will be in one place to be improved/fixed in a coherent way.
Reacted by 许杰友 Jieyou Xu (Joe), jyn and lolbinarycat
i would expect this to take cues from how cross-compilers already work, where we require specifically lldb-19 (or whatever) instead of just “the first lldb in PATH”
I'm surprised we don't already have a field in bootstrap.toml that allows you to specify an exact path for these like we do for npm.
We do, but if that field isn't set then we fall back to whatever's in PATH, which is typically what happens in practice.
(There doesn't seem to be a field for cdb specifically, despite compiletest having a --cdb command-line argument; I guess nobody ever fully hooked it up.)
Activity
Zalathar commentedon Nov 1, 2025
One of the follow-ups after #148099 will be to move the gdb and cdb discovery (which do the same thing) out into bootstrap, so that at least all of the questionable debugger discovery will be in one place to be improved/fixed in a coherent way.
jyn514 commentedon Nov 1, 2025
i would expect this to take cues from how cross-compilers already work, where we require specifically lldb-19 (or whatever) instead of just “the first lldb in PATH”
lolbinarycat commentedon Nov 5, 2025
I'm surprised we don't already have a field in
bootstrap.tomlthat allows you to specify an exact path for these like we do fornpm.Zalathar commentedon Nov 5, 2025
We do, but if that field isn't set then we fall back to whatever's in PATH, which is typically what happens in practice.
(There doesn't seem to be a field for
cdbspecifically, despite compiletest having a--cdbcommand-line argument; I guess nobody ever fully hooked it up.)jieyouxu commentedon Sep 7, 2026
Update: after #159455