On 1 December 2015 at 23:06, Richard Henderson <r...@twiddle.net> wrote: > On 12/01/2015 08:40 AM, Aurelien Jarno wrote: >> That said the problem reported is likely fixed by this commit that went >> just after it has been reported: > > > It does seem likely, but I don't see how we can know that the out size is > 1100 in that situation. The disassembler dump doesn't happen until after > we've done all of the writes that would have resulted in a highwater > overflow segv.
Yeah, if we always cleanly segv immediately on highwater overflow (as opposed to corrupting something so a later translation crashes) then this can't be the bug that's reported for qemu-i386. The actual TB that we never finish translating is quite small: IN: 0x419552e0: push %ebp 0x419552e1: mov %esp,%ebp 0x419552e3: sub $0x18,%esp 0x419552e6: fldl 0x8(%ebp) 0x419552e9: fstpl -0x8(%ebp) 0x419552ec: movl $0x14000000,0x4(%esp) 0x419552f4: movl $0x2,(%esp) 0x419552fb: call 0x41954b96 thanks -- PMM
