Instant PTSD. Only beaten by an error in, say, Line 84 in a file that has 72 lines only
Debugging Typescript without proper tools, in a nutshell. The original file might have 72 lines, but the transpiled JS file it’s debugging is like 3 times that many
That when the file isn’t minified, no shit it’s on line 1 there is only one line
Production errors at their finest
Honestly, you don’t want production to tell users what your exact errors are, and especially not what the source code looks like and where to find the problem exactly.
Production errors should mostly go “Well, this was on (y)our side, sorry about that” and nothing more.
If it’s in the minified front end code it’s already client side, of course you don’t show it to the user but they could find out if they wanted to. Server side errors are where you really have to watch out not to give out any details, but then logging them is also easier since it’s already on the server.
Ah you’re right, I’m thinking too much like a backend dev!
No, not all error have security implications
Not only that, but with JS you by definition have no code obfuscation. Minifying it just saves some bandwidth
Yeah, if one really wants obfuscation for JS, then it requires an extra build step. E.g. this: JavaScript-obfuscator
Enable sourcemaps
ugh… this reminds me of a compilation error I got that complained about an unknown character. Looked through the file, didnt notice anything, passed it around for other people to check and no one noticed anything until one person noticed that the quotes were being autocorrected to the straight quotes instead for some fucking reason. I didnt see it because 1) why would that even be a thing and 2) I am almost legally blind without my contacts.
Yep. I use a piece of software that lets me drag components onto a stage and will auto generate code in the back end. We can then add our own code to customize the components.
I ran into a super cryptic error message while working on a time sensitive project and I was ripping my hair out trying to figure out what the hell was wrong. Turns out someone had pasted some code in from an email and the custom code editor displays the straight quotes and angles quotes exactly the same but the back end code generation fortunately displays them differently. The program didn’t like the character set for the quotes used in Outlook but didn’t give me the proper line number, the error made zero sense, and looking at the code didn’t show that anything was wrong.
Learned the hard way that anything copied from a MS product gets pasted to a plain text editor first. I especially love how SharePoint sprinkles “zero with space” characters liberally throughout.
So I took a Marlin firmware for my 3d printer and just changed values like the bed size and added a BL Touch by uncommenting the specified line and the Visual studio errors out. I don’t get it
I’ve seen errors on line 0 in the past
Long inline assembler.
Might as well just delete it all and start again.
Even worse: “Error on line 1, column 8,653.”
Laughs in source map file.
My money is on a nonprinted char
Curse you, teams!!!
Sometimes I get this when I have unsaved changes to the file or the binary is not in sync with the code PEBKAC, and sometimes its the compiler 🙄
Yeah me too. It’s the compiler. 😏
See that ;? Take it out for dinner and it may tell you what’s wrong here
Try putting a
please;
statement in therealmost funny INTERCAL joke
Ha. I’ve just got a report of “UnknownError at foo in line 162”.
Line 162:
bar();
Of course, there’s a foo() nearby, and yeah the error is there. C# also has a problem with sourcemaps.
Anyway, the “UnknownError” is literal from the problem. And also, nothing has any problem at all, but foo() is from a proprietary 3rd party library…
deleted by creator
Kotlin is quite …special in this regard.
For certain language features, you have to declare functions as
inline
.
And when you do that, as far as I can tell, it basically just templates that function at the end of the files where that function is to be used.Because well, it suddenly starts to report line numbers in exceptions, which are beyond the end of the file. Which is not helpful, at all.
This can actually be caused by using mixed line endings. Which is a whole new class of horrible.
You should configure git to checkout and commit specific types of line endings.
“Sourcemaps, what’s that?”
Yes! Does anyone actually know why this happens!?
The error is usually with the line before
EDIT: could be a missing bracket:
if (x == 5){ do_thing();
or a comparison in place of an assignment:
x == 5;
It depends on what you’re using. I see it most often with TypeScript when the source maps are incorrect
When I forget a semicolon (in languages that like those), it gives the the error on the next line instead of the one that I forgot the semicolon on. It makes sense once you think about it but, man, it trips me up sometimes
Definitely also depends on the language. Here’s e.g. Rust, the goddamn overachiever, pointing arrows and everything:
Often feels like, if you know where it needs to go, why ask me?
IntelliJ has a setting like that, it’ll auto correct/lint/format files on save. Super useful, except sometimes when it’ll fuck up your entire file because it placed a bracket at the wrong place.
I’ve had it do that at some point when I forgot a semicolon in Javascript at exactly the wrong position. You see, Javascript can work without semicolons and it’ll automatically add them for you, except in a few cases where you need to be explicit.
For example:
const mishmash = 13 + 'world' [13].length
This will place the semicolon after the
.length
, turning the code intoconst mishmash = 13 + 'world'[13].length
.In my case, it tried to be smart about inserting a semicolon in an unexpected case, and it turned nested .map() call into something completely unexpected while I hit ctrl+a on unfinished code. Took me a git reset to get something readable out of the that file again!
As for Rust: using some macros tends to make it suggest baffling syntax alterations sometimes. When you fuck up a macro in such a way that the macro no longer generates correct syntax, you can get the Rust compiler to suggest placing semicolons halfway through statements or even words.
See, it’s this trying to be overly friendly nonsense I hate about JS. If you need semicolons, demand them. Don’t make it seem like you don’t then make your code break because it hudes that you do. My first orogramming job was at large multinational japanese motor company and they had a hard rule over no in house exe’s or opensource software. So the compromise was doing everything in JS. JS refused to listen to me on doing a single threaded for loop, just run the loop, wait a moment, run the next one, wait a moment…
JS, don’t help me, just do as I say
Doesn’t rust-analyser have the code suggestions that do fix it for you? It’s not fixing automatically, but it does know where it needs to go and it’s giving you a button that you click and it automatically fixes it
Oh yeah, it does. I’m not really complaining about fixing it myself. Mostly, I was joking that I felt like I’m unneeded. Rust-analyzer actually being able to fix it on its own, doesn’t help in that sense either. 🙃
Error: Unexpected the after the.
I assumed we weren’t talking “expected semicolon” since that ones pretty explanatory and would never appear on a blank line. That said, it does provide an opportunity to talk about the worst code style I ever saw. A dude decided he wanted semicolons at the start of lines so that compiler error always mapped to the line he would have put the semicolon on
At least there’s at least some benefit there :P
Some thoughts based on personal experience:
- File was edited after error was reported, but before investigating it. (Editing, code formatter, etc.)
- Build cache was not cleared, so old source was used.
deleted by creator
If you have optimizations enabled this can easily occur due to code being moved around behind the scenes. For example the body of a small function can be inserted where it is called. For example in c# if you get a null ref exception in such inlined code it will actually direct you to the place where the function is called from, rather than its body.
So if you know, you probably can find the issue. In the worst case just go to debug mode with opti off, or use an attribute on the function to force it not to inline.
Tears from async JS or transpiled without sourcemaps
:LspRestart
Haha preprocessor goes brrrrrrrrrrrrrrr