[Lldb-commits] [PATCH] D88728: [lldb] Check for and use ptsname_r if available

[Pavel Labath via Phabricator via lldb-commits]

labath added inline comments.


================
Comment at: lldb/source/Host/common/PseudoTerminal.cpp:149
+  int r = ptsname_r(m_primary_fd, buf, sizeof(buf));
+  assert(r == 0);
+  return buf;
----------------
MaskRay wrote:
> MaskRay wrote:
> > labath wrote:
> > > mgorny wrote:
> > > > labath wrote:
> > > > > mgorny wrote:
> > > > > > I would really feel better with a real error handling here. It shouldn't be hard to use `ErrorOr` here.
> > > > > Yeah, but what are you going to do with that value? Pass it to the caller? The existing callers are ignoring the error return anyway, and I don't want to add error handling everywhere as afaict, this function can't fail unless the user messes up the master state (which is not something I want to support).
> > > > I get your point but I've literally wasted days because of missing error handling, so I'd really preferred if we wouldn't make it even worse. Though I guess `assert` is good enough.
> > > In some ways it's even better because it will point you straight to the place where the assumption is violated, whereas a propagated logic error can manifest itself much farther away (or not at all). :)
> > If `ptsname/ptsname_r` fails, buf will be uninitialized and trigger a use-of-uninitialized-value error.
> ... in a -DLLVM_ENABLE_ASSERTIONS=off build.
> 
> This probably still needs some protection.
What kind of protection did you have it mind? Initialize the buffer to an empty string?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88728/new/

https://reviews.llvm.org/D88728