7 Jun
2016
7 Jun
'16
8:39 a.m.
On Fri, Jun 3, 2016 at 6:28 PM, Waldemar Brodkorb wbx@uclibc-ng.org wrote:
If you provide one patch for each architecture I can test them and if no regression happens I can commit them.
Excellent! Patches attached. The single platform that keep local lowlevellock.c is sparc, like glibc (sysdeps/sparc/sparc32/lowlevellock.c).
Do you think our testsuite covers the included functionality in lowlevellock.c enough, so that no failure count changes, mean the generic code is fine?
Testsuite cover standard, predictable, situations. But I'm unsure about races & other unusual problems tests. Yes, this adds some potential risk, but I haven't any good idea how to eliminate it at all.
regards, Leonid