Smalltalk › Squeak › Squeak VM

Difficult to debug VM crash with full blocks and Sista V1

‹ Previous Topic Next Topic ›
 
classic Classic list List threaded Threaded
♦
♦
 24 messages Options
Eliot Miranda-2 Difficult to debug VM crash with full blocks and Sista V1 – Hi All, there is a VM bug in 64-bit Spur with the Sista V1 bytecode set and full blocks. The symptom is that when waiting for a remote ...
Jan Vrany Hi, > > - can anyone think of any other strategies I might take to try and > reproduce this? > at the risk of stating the obviou...
Nicolas Cellier Hi Jan, Thanks, a recording debugger would indeed be a tremendous help! Unfortunately, I'm unable to reproduce the bug under linux, only macos...
Jan Vrany On 2019-09-22 10:57, Nicolas Cellier wrote: > Hi Jan, > Thanks, a recording debugger would indeed be a tremendous help! > Unfortunately...
Nicolas Cellier Follow up: so we have a problem of repeatability... I found nothing like rr for macos, and CPU watchpoint capability is too restricted to be u...
Nicolas Cellier Le mar. 24 sept. 2019 à 00:25, Nicolas Cellier < nicolas.cellier.aka.nice@...> a écrit : > Follow up: so we have a problem of repea...
Nicolas Cellier I noticed that many crash happen in testSendTimeout. This test is organizing a race: - Smalltalk fill the socket send buffer - the OS tries t...
Eliot Miranda-2 Hi Nicolas, On Wed, Sep 25, 2019 at 12:32 AM Nicolas Cellier < nicolas.cellier.aka.nice@...> wrote: > > I noticed that many ...
Nicolas Cellier Hi Eliot, so I have set watchpoint to 4 previous crash and got this: (lldb) run ../../image/crash.image There is a running process, kill it...
Nicolas Cellier Hi all, So I finally got it! That's our specific handling of SIGIO with forceInterruptCheck. Don't you JIT the signal handler! The signal ...
Nicolas Cellier Maybe I draw conclusions too quickly. But clearly, the signal handler has a stack pointer pointing in methodZone. (lldb) p/x $rsp (unsigned...
Nicolas Cellier I confirm that each crash follows a SIGIO -> forceInterruptCheck() with corrupted stack pointer pointing to methodZone (that is executable z...
Nicolas Cellier If I comment /* signal(SIGIO,forceInterruptCheck) */ in platforms//unix/vm/aio.c then I don't experiment crashes anymore One solution if we...
Eliot Miranda-2 On Thu, Oct 3, 2019 at 1:47 PM Nicolas Cellier < nicolas.cellier.aka.nice@...> wrote: > If I comment /* signal(SIGIO,forceInterrupt...
David T. Lewis Hi Nicolas, I am only reading your (very impressive) emails here, and not really looking at the code. But for whatever it may be worth: I am n...
Eliot Miranda-2 Hi Nicolas, On Thu, Oct 3, 2019 at 1:10 PM Nicolas Cellier < nicolas.cellier.aka.nice@...> wrote: > I confirm that each crash fo...
Ben Coman On Wed, 25 Sep 2019 at 15:32, Nicolas Cellier < nicolas.cellier.aka.nice@...> wrote: > > I noticed that many crash happen in te...
Nicolas Cellier Thanks Ben, nice references! I opted for a simple reduction of TestCase timeout and/or Smalltalk buffer size, and could easily obtain a TestCa...
Eliot Miranda-2 Hi Jan, this could be very helpful. Care to hold my hand? Does it work on macOS? (we're only seeing the crash on macOS, and then only...
Jan Vrany On Sun, 2019-09-29 at 15:49 -0700, Eliot Miranda wrote: > > Hi Jan, > > this could be very helpful. Care to hold my hand? ...
Eliot Miranda-2 Hi All, two further reports. One is that attempting to simulate SocketTest suite run with a small code zone to provoke code comp-action...
Nicolas Cellier Hi Eliot, thanks for the heads up. Also remember that I was not able to produce such crash on Windows nor linux spur64, and I don't remember o...
Nicolas Cellier So there is compaction of cog method zone of course... Now that I have a VMMaker image under the eyes it's clear: cogit compactCogCompiledCode...
Phil B I'm sure you're right (assuming the GUI tool still exists... the top answer is quite old.) Mainly I linked to the answer since it seems to cov...
Free forum by Nabble Edit this page