On 19/10/20 19:16, Emmanuel Blot wrote: > --- undef_sh.log 2020-10-19 19:15:08.000000000 +0200 > +++ undef_py.log 2020-10-19 19:15:20.000000000 +0200 > @@ -1,3 +1,4 @@ > +-Wl,-u, > -Wl,-u,_PreallocMode_lookup > -Wl,-u,__TRACE_CURL_CLOSE_DSTATE > -Wl,-u,__TRACE_CURL_OPEN_DSTATE > > > I think it can be addressed with the following patch: > > From d157bd508d32485dd5cdb3efcff7969a89628cc1 Mon Sep 17 00:00:00 2001 > From: Emmanuel Blot <emmanuel.b...@sifive.com> > Date: Mon, 19 Oct 2020 19:07:03 +0200 > Subject: [PATCH] Fix regression with undefined symbol filter introduced in > 604f3e4e. > > --- > scripts/undefsym.py | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/scripts/undefsym.py b/scripts/undefsym.py > index 69a895cd26..254c456f25 100644 > --- a/scripts/undefsym.py > +++ b/scripts/undefsym.py > @@ -21,6 +21,8 @@ def filter_lines_set(stdout, from_staticlib): > continue > if not from_staticlib and tokens[1] != b'U': > continue > + if not tokens[0]: > + continue > new_line = b'-Wl,-u,' + tokens[0] > if not new_line in linesSet: > linesSet.add(new_line)
Great, thanks. Can you check if my patch would also work, as it would adhere more to the pre-conversion logic? Paolo