On Fri, 2019-03-01 at 07:04 -0800, Samuel Thibault wrote: > Svante Signell, le lun. 25 févr. 2019 10:25:01 +0100, a ecrit: > > Unfortunately, the situation has not improved with it :( > > > > The SIGILL failures are now SIGABRT and SIGSEGV as they became with the > > SS_DISABLE patch. > > It however seems that the failures are different? > (possibly with much more understandable backtraces)
No they are not from a global tests perspective. That requires further investigations. I suspect the problems arise earlier in the backtrace chain. One suspicion is the *context functions. Can they be called more than once without problems? Unfortunately I don't know anything (have forgotten everything) about assembly code... Just a thought!