SpotLess tests need it to find out one bug

	Bottom Previous Topic Next Topic
Register To Post

« 1 (2) 3 4 5 »

jabirulo

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 12:50 #21

Just can't stay away

@kas1e

You mean the crash with de TEST executable you posted?

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 13:33 #22

Home away from home

@javier
Yeah, mean that test i uploaded, as we use it as one cause issues

Join us to improve dopus5!
AmigaOS4 on youtube

jabirulo

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 14:09 #23

Just can't stay away

@kas1e

tested a few times and stops/works ok unitl I continue from the breakpoint on puts():


actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

actions->update() BEGUN

actions->update() DoNE

memory_insert_break_instruction : 0x7d31e228

actions->update() BEGUN

actions->update() DoNE

memory_remove_break_instruction : 0x7d31e228

actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

memory_insert_break_instruction : 0x7d31e22c

memory_remove_break_instruction : 0x7d31e22c

memory_insert_break_instruction : 0x7d31e228

memory_insert_break_instruction : 0x7d31e214

memory_insert_break_instruction : 0x7d31e228

memory_insert_break_instruction : 0x7d31e234

memory_insert_break_instruction : 0x7d31e240

memory_insert_break_instruction : 0x7d31e244

actions->update() BEGUN

actions->update() DoNE

memory_remove_break_instruction : 0x7d31e214

memory_remove_break_instruction : 0x7d31e228

memory_remove_break_instruction : 0x7d31e234

memory_remove_break_instruction : 0x7d31e240

memory_remove_break_instruction : 0x7d31e244

memory_remove_break_instruction : 0x7d31e228

actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

but when recompiling TEST.C and it works fine, so I thinks is a "buggy" test executable.
Recompiled TEST.C (see I get 'Hello' in output and nothing on above "buggy" test)


actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

actions->update() BEGUN

actions->update() DoNE

memory_insert_break_instruction : 0x7fe4648c

actions->update() BEGUN

actions->update() DoNE

memory_remove_break_instruction : 0x7fe4648c

actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

memory_insert_break_instruction : 0x7fe46490

memory_remove_break_instruction : 0x7fe46490

memory_insert_break_instruction : 0x7fe4648c

memory_insert_break_instruction : 0x7fe46478

memory_insert_break_instruction : 0x7fe4648c

memory_insert_break_instruction : 0x7fe46498

memory_insert_break_instruction : 0x7fe464a4

memory_insert_break_instruction : 0x7fe464a8

actions->update() BEGUN

Hello.



actions->update() DoNE

memory_remove_break_instruction : 0x7fe46478

memory_remove_break_instruction : 0x7fe4648c

memory_remove_break_instruction : 0x7fe46498

memory_remove_break_instruction : 0x7fe464a4

memory_remove_break_instruction : 0x7fe464a8

memory_remove_break_instruction : 0x7fe4648c

actions->update();

actions->update() BEGUN

actions->update() DoNE

code->update();

context->update();

stacktrace->update();

disassembler->update();

if(memorySurfer && memorySurfer->windowObject()) memorySurfer->update();

updateAll() DoNE

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 14:11 #24

Just can't stay away

@thread

Ok, so I might explain. The idea is, that perhaps the reading of context info on a running child could be responsible for the crashes. Usually, if you have a shared ressource and multiple reads at the same time, it will screw up. Also, it might give bad results, even if it succeeds. I think, I might have closed one of the gaps for this scenario. There might be a small case left. If I can manage to close this, my hope is, that we will have solved this crash scenario.

I just came back from the hospital with a lung problem. Now I am in a summer house with some friends. When I get back Saturday, I might be able to do a fresh version. But of course I can't test it myself, since my x5000 is dead

walkero

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 14:37 #25

Site Builder

@alfkil
Quote:

I just came back from the hospital with a lung problem.

Take care of yourself and get well soon.

Quote:

But of course I can't test it myself, since my x5000 is dead

Is it completely dead? Do you know if it is something that you could bring back to life?
I would recommend a separate thread to help you with that, as much as possible.

Follow me on
Ko-fi, Twitter, YouTube, Twitch

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 15:01 #26

Just can't stay away

@walkero

I am quite sure, that it is a fatal wound (the x5000). I am communicating with Amigakit, but apparently they haven't got so many options at the moment.

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 16:00 #27

Home away from home

@alfkil
Quote:

Ok, so I might explain. The idea is, that perhaps the reading of context info on a running child could be responsible for the crashes. Usually, if you have a shared ressource and multiple reads at the same time, it will screw up. Also, it might give bad results, even if it succeeds. I think, I might have closed one of the gaps for this scenario. There might be a small case left. If I can manage to close this, my hope is, that we will have solved this crash scenario.

Just to clarify : new version do not fix bug we have with test case we use (and javier cofirm it). But the same recompiled source do support it.

What make us think that test case broken, but! it loads up, it works, breakpoints sets, so we need to understand what "broken" in, to avoid crashes in debugger.

Join us to improve dopus5!
AmigaOS4 on youtube

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 16:49 #28

Just can't stay away

@kas1e

But before you said, that there was some improvement?? If there is no improvement with this version, then I have no idea what to do.

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 18:15 #29

Home away from home

@alfkil
Yeah, improvements there for sure, i can't it crash that easy as before. But i was under impression that you didn't change anything, but only add more "prinfs" to console and i was under impression that was reasson for shifting issue. But if you add something else, then yeah, it can mean that things changes.

@Javier
Can you confirm that it harder now to reproduce bug on our "buggy" test case ?

Join us to improve dopus5!
AmigaOS4 on youtube

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 19:27 #30

Just can't stay away

@kas1e

I did change something, and was expecting it to perform Better. I have one last change, I want to do, and then remove the console output. If we are lucky, this is going to do the trick.

jabirulo

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/6 21:03 #31

Just can't stay away

@kas1e

ok will test a few more tines times the "buggy" test executable and post here result.
BTW my system is SAM460ex, so it can be different from X50x0.

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/9 17:14 #32

Just can't stay away

New test candidate : Spotless test alpha 02

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/9 17:53 #33

Home away from home

@Alfkil
Tried 5 times in a row this "bad" test case : seems all fine now. But i also see the prinfs in console as well, so hope it's not them shift the deadlock :)

Btw, do you have any amigang hw instead of x5000 ? sam460 or something ? I mean how to developer debugger futher without real amiga hw

Join us to improve dopus5!
AmigaOS4 on youtube

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/9 18:19 #34

Just can't stay away

@kas1e

Quote:

Tried 5 times in a row this "bad" test case : seems all fine now. But i also see the prinfs in console as well, so hope it's not them shift the deadlock :)

Are you serious??! Whupti-woo!

No, it is not because of the prints. It is because I understand, what is wrong :).

I will do a freshly cleaned version in just a bit. Sadly I have no other Amiga hardware. I am trying to repair it with a new power supply together with a friend. If unsuccessfully, I will have to buy new hardware.

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/10 12:42 #35

Just can't stay away

Fresh version with a registers pane. Untested.

Spotless without prints and with registers pane

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/10 16:10 #36

Home away from home

@Alfkil
Tested registers part. From the first look noticed that all fpr ones have some madness values like:

fpr[0] : -1188595570704978439862-2380927392387239872387923072307230723072307923709237020370273270323072307 and so on and so on. Even didn't fit into the window.

That just when i simple load test case and do nothing. Only load it. Then hit breakpoint in our "bad" test case on puts("hello"), and registers updates, etc, as expected. Most of FPR registers start to show "sane" values, but not all.

Question is : should't we "clear" all registers when we simple do "load" of binary, without atually hitting a start ?

Next, find out again crash. 100% reproducable all the time. To reproduce:

-- load our test case
-- hit bp on "puts("hello")
-- hit start (so it breaks)
-- hit "step out" button, and immediately crashes. Stack trace are:


Stack trace:

(0x619F7460) [/home/alfkil/Spotless-ahead/SimpleDebug/Breaks.cpp:155] Spotless (46):_ZN6Breaks31memory_insert_break_instructionEjPj()+0xcc (section 1 @ 0x50F58)

(0x619F7490) [/home/alfkil/Spotless-ahead/SimpleDebug/Breaks.cpp:152] Spotless (46):_ZN6Breaks31memory_insert_break_instructionEjPj()+0xa8 (section 1 @ 0x50F34)

(0x619F74C0) [/home/alfkil/Spotless-ahead/Spotless/Spotless.cpp:274] Spotless (46):_ZN8Debugger7stepOutEv()+0xb8 (section 1 @ 0x9900)

(0x619F74E0) [/home/alfkil/Spotless-ahead/Spotless/Spotless.cpp:115] Spotless (46):_ZN7Actions11handleEventEP5Event()+0x4f0 (section 1 @ 0x121C8)

(0x619F7710) [/home/alfkil/Spotless-ahead/ReAction/Widget.cpp:403] Spotless (46):_ZN6Widget12processEventEmt()+0x728 (section 1 @ 0x2BB0C)

(0x619F7900) [/home/alfkil/Spotless-ahead/ReAction/Widget.cpp:194] Spotless (46):_ZN6Widget12waitForCloseEv()+0x594 (section 1 @ 0x2A92C)

(0x619F7AB0) [/home/alfkil/Spotless-ahead/Spotless/Spotless.cpp:63] Spotless (46):_ZN8Spotless6unfoldEv()+0x54 (section 1 @ 0x600C)

(0x619F7AD0) [/home/alfkil/Spotless-ahead/main.cpp:5] Spotless (46):main()+0x114 (section 1 @ 0x339C)

(0x619F7D20) native kernel module newlib.library.kmod+0x00002614

(0x619F7D70) native kernel module newlib.library.kmod+0x00003340

(0x619F7F20) native kernel module newlib.library.kmod+0x00003864

(0x619F7F50) Spotless (46):_start()+0x170 (section 1 @ 0x3210)

(0x619F7F90) native kernel module dos.library.kmod+0x0002a490

(0x619F7FC0) native kernel module kernel+0x0005ede4

(0x619F7FD0) native kernel module kernel+0x0005ee5c

PS. Imho our current GUI layout start to act up. I mean, you need to click over the tabs all the times to see disassembler and registers. Should't be they separate windowses, or sub-windowses , which we always see ?

EDIT:

Also found GUI bug, to reproduce:

-- run spotless
-- hit RMB , choose "configure"
-- see, on the left/top side, you have some empty small square appears and stay there since that time forever.

EDIT2:

Another issue/sometime-crash:

-- run spotless
-- hit load , choose our test case, hit "bp" on "puts hello"
-- hit start. once break, start to hit "step over" until program ends and exit, and load button again will colorized

-- load again same test case, hit "bp" on "puts hello"
-- hit start. once break, now start to hit "step into" until program ends and exit, and load button again will be colorized.

-- now, load again test case, hit again "bp" on "puts hello" and hit start.

At this point, you or crashes, or disassembler window will show you <not a readable address>, or all other sort of issues.

Edited by kas1e on 2022/4/10 16:26:32
Edited by kas1e on 2022/4/10 16:27:32

Join us to improve dopus5!
AmigaOS4 on youtube

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/4/11 14:45 #37

Just can't stay away

@kas1e

Thanks for the reports. As you might guess, I cannot do so much at the moment. I will save info and work when possible.

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/5/16 15:53 #38

Just can't stay away

@kas1e

This is my attempt at fixing first two issues from your post. Last two, I am not sure what to do about.

Spotless : Download from Drive

kas1e

Re: SpotLess tests need it to find out one bug

Posted on: 2022/5/16 18:06 #39

Home away from home

@alfkil
Thanks! Will check.

Do you fix your x5k ?

Join us to improve dopus5!
AmigaOS4 on youtube

elfpipe

Re: SpotLess tests need it to find out one bug

Posted on: 2022/5/16 18:24 #40

Just can't stay away

@kas1e

A friend of mine picked up a new power supply from the mail. We will try to install it asap.

Register To Post	« 1 (2) 3 4 5 »
	Top Previous Topic Next Topic

Currently Active Users Viewing This Thread: 2 ( 0 members and 2 Anonymous Users )