My app seems to be leaking some allocations, at least instruments says so, but it points me to the wrong line of code.
The stack trace in the image bellow, says that btAllignedAllocDefault is leaking, but it is just a fancy malloc, so the offending line of code must be in initEngine, the problem is that double-ckicling initEngine shows the file containing iniEngine's definition but always point to some arbitrary line in the first function definition it finds in that file.
No mather where I put the supposed offending function initEngine(), Instruments points to some arbitrary line of code always in the first function it finds in the file that contains the offendig code.
I tryied deleting all the DerivedData, and building both the Release and Debug build configurations.
I've noticed, though, that Xcode never puts the .app bundle and .dSYM file in Debug-iphoneos, but always in Release-iphoneos, even when the Build location settings tell it to do so.
Related
Scenario:
create a .gd script (i.e. in GDScript)
Make it so that the script outputs an error to the Godot console.
More specifically: Make it so that the error is not a crash inside the low-level c++ code, but strictly something in the GDscript. For example, this kind of instruction: assert(false) .
This does not need to involve c++ in any way. Therefore in my opinion the location of this game error in the c++ codebase is irrelevant. All I want to know is in which GDScript this happened.
Yet, Godot says in which cpp file the error "occurred". Every time, it creates an extra line in the logs.
It slows down my work a lot: Every time, I need to read twice: First my brain makes me read that incorrect location (muscle memory: if the logs give you a stack trace, you look at the stack trace!). It takes one or two seconds for me to discard that line and read what's above: The "real" error with the "real" location of the problem (the GDscript).
Is there a way to not display the cpp file location whenever an error occurs in a GDScript?
It's not possible at the moment.
Repo for all code I've been using is updated here . When I run the requestor script it exits with a runtime error 2 (File not found). I am not sure how to further debug this or fix it. So far I've converted my code over to a python slim docker image to better mirror the example. It also works for me when I spin up a docker image that typing and running "/golem/work/imageclassifier.py --trainmodel" works from root. I switched all my code to use absolute paths. I also did make sure the shebang (#!) uses linux end of file characters rather than windows before which was giving me errors. Fixed a bug where my script returns error code 2 when called with no args to now pass.
clf.fit(trainDataGlobal, trainLabelsGlobal)
pkl_file = "classifier.pkl"
with open(pkl_file, 'wb') as file:
pickle.dump(clf, file)
is the only piece I could think of that causes the issue, but as far as I can tell this is the proper way to pickle something in python. Requestor script is also heavily based on the simple service example and I tried to mirror my design to that. I just need help in getting more information while debugging, or guidance on how to move forward from here
So Flow only works correctly the first time I run it, and then I have to restart my computer before it'll work correctly again.
Specifically, the problem I'm seeing is that we are using the Flow language to add type annotations to our JS code. Our linter script is setup to run flow type checking among other things. However, when I fix an issue in my code and then rerun the linter script, it still comes back with the exact same errors... BUT when it shows the piece of code where the error is supposed to be, it actually shows my updated code that's fixed.
So as an example, I had a file I copied into the project, that I didn't think I really needed, but maybe I would. So I copied it in just in case. Well then it came up with a bunch of linter errors, so I decided to just delete the file since I didn't really need it. So then I run "yarn lint --fix" again, but it's still complaining about that file, EVEN THOUGH THE FILE DOESN"T EXIST! Now interestingly, where the linter output is supposed to show the code for those errors it's just blank.
Or another example, let's say I had a couple of functions in my code:
100: function foo() {}
...
150: function bar() {}
And foo has a lot of errors because it was some throw away code I don't need anymore and so I just delete it. So the new code looks like:
100: function bar() {}
Well I rerun the linter and get an error like:
Error ------------------------ function foo has incorrect
something...blah blah
src/.../file.js
100| function bar() {}
I also tested this out on a coworker's machine and they got the same behavior that I did. So it's not something specific to my machine, although it could be specific to our project?
Note: There doesn't appear to be a tag for Flow, but I couldn't post without including at least one tag, so I used flowlang even though that's actually a different language :-( I'm assuming that anyone looking for flow would also use that tag since it's the closest.
The first time you launch Flow it starts up a background process that is then used for subsequent type checking. Unfortunately this background process is extremely slow, and buggy to boot. In linux you can run:
killall flow
To stop the background process. Then if you rerun the flow type checker, it will actually see all your latest changes.
I have an app that has been running fine on the iPhone simulator for some time. Recently, I decided I wanted to re-use the data model and related classes in another project - so I dragged them from this project window to the other then told Xcode not to copy, just to make references. At first this didn't work so I jumped through a number of hoops to try to fix it (I may be asking more about that in another post). After all this, I re-compiled and tried to run the original app -- and it's not working any more. On further investigation, I discovered that when I re-compile the original app, I end up with a bundle that contains a .momd package but it contains only a Versioninfo.plist file - no .mom file, no .omo file like I'm expecting to see. I don't recall making any changes to the original app. I don't get any warnings. I just get an incomplete .momd package (and, not surprisingly, my app now crashes).
What's going on here?
BTW, the app now crashes with this message:
Terminating app due to uncaught exception 'NSInvalidArgumentException', reason: '* -[__NSArrayM insertObject:atIndex:]: object cannot be nil'
Which I get when executing this line of code:
self.productRegistry = [[UIManagedDocument alloc] initWithFileURL:self.productRegistryURL];
I figured this out by looking more closely at the file locations in the project directory using Finder. In the Xcode window, everything looks normal but in the actual project directory I found that the .datamodeld package had ended up at the top level of the project directory -- at the same level as the project package itself. Xcode apparently did not like this but unfortunately it did not complain -- it just created a partial build output. Once I moved the .datamodeld package into the same folder as the rest of the project's code, everything worked just fine.
This would appear to be just a quirk. I would expect that Xcode would either see that all is well and build correctly OR it would see that things weren't quite as they should be and fail. In this case, it did not build correctly but was silent about it.
Hope this answer helps someone else someday.
The Map file looks like:
0002:000442e4 00000118H .idata$2 DATA
0002:000443fc 00000014H .idata$3 DATA
0002:00044410 00000b7cH .idata$4 DATA
0002:00044f8c 0000512eH .idata$6 DATA
0002:0004a0ba 00000000H .edata DATA
The Crash info looks like:
Application Error : The instruction at "0x00458ae1" referenced memory at "0x00000074". The memory could not be "read".
I'm trying to get a stack dump on the next crash, but it seems to me this is a case where we trounced the stack, then did a return, which made us end up executing data.
I'm not entirely certain though because I read some articles like this: Under the Hood Article seems to indicate this is an area of imported method names
The data that an import library provides for an imported API is kept
in several sections whose names all begin with .idata (for instance,
.idata$4, .idata$5, and .idata$6). The .idata$5 section contains a
single DWORD that, when the executable loads, contains the address of
the imported function. The .idata$6 section (if present) contains the
name of the imported function. When loading the executable into
memory, the Win32 loader uses this string to call GetProcAddress on
the imported function effectively.
Without a stack backtrace I'm kind of stuck. Am I looking at this crash the wrong way?
Forget MAP files, better use PDB files. For this enable linker option /DEBUG - yes, even for Release builds. /DEBUG is linker option, _DEBUG is compiler option. Only _DEBUG controls the code, and any conditional compilation that source/headers have put against this.
Debug builds have optimizations disabled, _DEBUG macro enabled.
Release builds have optimizations enabled, _DEBUG macro disabled.
/DEBUG would just put debugging-information into the EXE/DLL, and wont affect anything else.
Coming back to problem, when crash occurs. Do NOT close the application when WER (Windows Error Reporting) says it crashed. Instead keep it there, goto Task Manager, goto Process tab, select that crashed/crashing process, and hit "Create Dump File". Dump file (full-dump) will be created in some local folder (the path will be shown by task-manager). You can now close the crashing application (the WER window).
Now copy this .DMP file into some safe location, preferably the folder having your original Release folder. Open it in Visual Studio or WinDbg. On VS, just hit F11/F10, and you will be shown call stack. If multiple threads are running (in your crashed application), launch "Threads" view, and see the only suspended thread, double click it and you'll find the crash location.
You must have correct PDBs along with all binaries, and absolutely same code to see Code, otherwise call stack wont be good.
To get more information about PDB and stuff, you can read this article.