Pharo 6, bad debugger behaviour
Hi, Iâve noticed lately when using a reasonably recent Pharo 6.1 setup that if I restart a method in the debugger - particularly from a breakpoint or correcting a dnu issue the environment frequently locks up. Fearing the worst (after a few seconds), if I repeatedly press cmd . (Say 5 times) after 10 seconds more it comes back with 30 debugger Windows typically with something like Instance of Xxxx dnu: #someMethod (which is legit as I might have a now corrupt inst bar). I understand the error, but why does it lockup and require the cmd . Dance? I dont recall earlier Pharos doing this (and not other smalltalks). Should I report this? Tim Sent from my iPhone
I should also mention these extra debuggers donât have the stack either, just a single dnu method. Tim Ps thank god for close windows to right. ... Sent from my iPhone
On 8 Jun 2018, at 12:30, Tim Mackinnon <tim@testit.works> wrote:
Hi, Iâve noticed lately when using a reasonably recent Pharo 6.1 setup that if I restart a method in the debugger - particularly from a breakpoint or correcting a dnu issue the environment frequently locks up. Fearing the worst (after a few seconds), if I repeatedly press cmd . (Say 5 times) after 10 seconds more it comes back with 30 debugger Windows typically with something like Instance of Xxxx dnu: #someMethod (which is legit as I might have a now corrupt inst bar).
I understand the error, but why does it lockup and require the cmd . Dance?
I dont recall earlier Pharos doing this (and not other smalltalks).
Should I report this?
Tim
Sent from my iPhone
Am 08.06.2018 um 13:30 schrieb Tim Mackinnon <tim@testit.works>:
Hi, Iâve noticed lately when using a reasonably recent Pharo 6.1 setup that if I restart a method in the debugger - particularly from a breakpoint or correcting a dnu issue the environment frequently locks up. Fearing the worst (after a few seconds), if I repeatedly press cmd . (Say 5 times) after 10 seconds more it comes back with 30 debugger Windows typically with something like Instance of Xxxx dnu: #someMethod (which is legit as I might have a now corrupt inst bar).
I understand the error, but why does it lockup and require the cmd . Dance?
I dont recall earlier Pharos doing this (and not other smalltalks).
Should I report this?
As an exercise for you. If you go to http://bugs.pharo.org <http://bugs.pharo.org/> and enter âdebuggerâ in the search field, please click on the first entry found which should be 22085. Norbert
On 08/06/2018 08:30, Tim Mackinnon wrote:
Hi, Iâve noticed lately when using a reasonably recent Pharo 6.1 setup that if I restart a method in the debugger - particularly from a breakpoint or correcting a dnu issue the environment frequently locks up. Fearing the worst (after a few seconds), if I repeatedly press cmd . (Say 5 times) after 10 seconds more it comes back with 30 debugger Windows typically with something like Instance of Xxxx dnu: #someMethod (which is legit as I might have a now corrupt inst bar).
I understand the error, but why does it lockup and require the cmd . Dance?
I dont recall earlier Pharos doing this (and not other smalltalks).
Should I report this?
It was reported years ago [1], there was a bug and an issue, and it was supposedly fixed, apparently it wasn't. In some cases that recursion caused a "detachment" of the image from the changes, it is, the image starts complaining that it can't find the changes file, so all methods lose the source and you better discard that image without saving, because if you do, you'll never be able to "reattach" it. I don't know if it is still around in Pharo 7. Regards, [1] https://twitter.com/emaringolo/status/603939752098816000 -- Esteban A. Maringolo
Good point by Norbert - I should have referenced a bug - but interested in others experience and I will annotate that bug. For me, this isnât a recursion thing, itâs a dnu when restarting a method (not sure if many people do this - but itâs a key Smalltalk feature in my mind - so it should work well). Also my image recovers fine when I shut down the debuggers. Interested in others experience and I guess I should try and use 7 more to see if itâs there too. Tim Sent from my iPhone Sent from my iPhone
On 8 Jun 2018, at 13:10, Esteban A. Maringolo <emaringolo@gmail.com> wrote:
On 08/06/2018 08:30, Tim Mackinnon wrote: Hi, Iâve noticed lately when using a reasonably recent Pharo 6.1 setup that if I restart a method in the debugger - particularly from a breakpoint or correcting a dnu issue the environment frequently locks up. Fearing the worst (after a few seconds), if I repeatedly press cmd . (Say 5 times) after 10 seconds more it comes back with 30 debugger Windows typically with something like Instance of Xxxx dnu: #someMethod (which is legit as I might have a now corrupt inst bar).
I understand the error, but why does it lockup and require the cmd . Dance?
I dont recall earlier Pharos doing this (and not other smalltalks).
Should I report this?
It was reported years ago [1], there was a bug and an issue, and it was supposedly fixed, apparently it wasn't.
In some cases that recursion caused a "detachment" of the image from the changes, it is, the image starts complaining that it can't find the changes file, so all methods lose the source and you better discard that image without saving, because if you do, you'll never be able to "reattach" it.
I don't know if it is still around in Pharo 7.
Regards,
[1] https://twitter.com/emaringolo/status/603939752098816000
-- Esteban A. Maringolo
Concerning the bug 22085, i commented on fogbuzz:
I believe it is because when running test cases, it runs through:
runCaseForDebug: aTestCase [...call test code...] on: self class failure , self class skip, self class warning, self class error do: [:ex | ex sunitAnnounce: aTestCase toResult: self. ex pass]
So when you reach a doesNotUnderstand, you have:
doesNotUnderstand: aMessage [...stuff...]
resumeValue := exception signal.
^exception reachedDefaultHandler ifTrue: [aMessage sentTo: self] ifFalse: [resumeValue]
In normal cases (class of the receiver is not a test case), when you step in the debugger the exception signal blocks the execution right away. In subclasses of TestCase, it does ex pass each time you get to this line, then it enters the aMessage sentTo: self, which is not understood, then you can wait a long time...
It is equivalent to doing this in a playground:
[ Object new doStuff ] on: Exception do:[:ex| ex pass]
Then step in the debugger and bye bye. But then it seems to be expected behavior... or maybe not ? I mean voluntarily passing on an exception in a doesNotUnderstand would lead to such problems.
I have heard at least three different stories about this bug : yours, the one from fogbuzz and one other a bit different from Guille. They seem to be different but all of them end up looping ad vitam eternam after a debugger step. I am looking for ways to reproduce any of this bugs (well except 22085), but so far i have no success on my own images :'( (pharo 6.1 & 7). Steven.
Ah thatâs interesting- itâs happened to me 4 or 5 tines in the last 2 days. Recent enough that maybe I can figure out a test case.... Iâll try. Sent from my iPhone
On 8 Jun 2018, at 18:57, Steven Costiou <steven.costiou@kloum.io> wrote:
Concerning the bug 22085, i commented on fogbuzz:
I believe it is because when running test cases, it runs through:
runCaseForDebug: aTestCase [...call test code...] on: self class failure , self class skip, self class warning, self class error do: [:ex | ex sunitAnnounce: aTestCase toResult: self. ex pass]
So when you reach a doesNotUnderstand, you have:
doesNotUnderstand: aMessage [...stuff...]
resumeValue := exception signal.
^exception reachedDefaultHandler ifTrue: [aMessage sentTo: self] ifFalse: [resumeValue]
In normal cases (class of the receiver is not a test case), when you step in the debugger the exception signal blocks the execution right away. In subclasses of TestCase, it does ex pass each time you get to this line, then it enters the aMessage sentTo: self, which is not understood, then you can wait a long time...
It is equivalent to doing this in a playground:
[ Object new doStuff ] on: Exception do:[:ex| ex pass]
Then step in the debugger and bye bye. But then it seems to be expected behavior... or maybe not ? I mean voluntarily passing on an exception in a doesNotUnderstand would lead to such problems.
I have heard at least three different stories about this bug : yours, the one from fogbuzz and one other a bit different from Guille.
They seem to be different but all of them end up looping ad vitam eternam after a debugger step.
I am looking for ways to reproduce any of this bugs (well except 22085), but so far i have no success on my own images :'( (pharo 6.1 & 7).
Steven.
participants (4)
-
Esteban A. Maringolo -
Norbert Hartl -
Steven Costiou -
Tim Mackinnon