feat(gstd): get panic message without panic_info_message

[StackOverflowExcept1on]

Resolves #3524

We are getting closer to cargo build +stable -p gstd

@gear-tech/dev

[breathx]

Still considering best approach here

[breathx]

So what have we decided to do with possibility to throw whole display for not(feature(debug))? @StackOverflowExcept1on
And how is it working wout oom-handler and allocation error on stable panic message? It would be good to have some tests maybe

[StackOverflowExcept1on]

@breathx

So what have we decided to do with possibility to throw whole display for not(feature(debug))?

I think it's quite expensive even for the nightly version. core::fmt::write uses about 2.5 KiB, and extra 2.5 KiB is used for impl Display for u32 (it's quite large). Thus, contracts will increase by 5 KiB on the stable version and by 2,5 KiB on nightly.

How is it working wout oom-handler and allocation error on stable panic message?

We've already discussed this. On the nightly version, gr_oom_panic syscall is called. On the stable version, gr_panic syscall is called with message like this: 'memory allocation of 65536 bytes failed', library/alloc/src/alloc.rs:411:13. This creates bit difference in behavior, but in general it is unlikely that anyone will rely on it in tests.

[breathx]

Got your point. I'm good with the solution. Awesome work
Please review @ark0f

[ark0f]

I'd just print PanicInfo without any parsing. Do we have cons with this approach?

[StackOverflowExcept1on]

@ark0f there are several reasons:

1. As user of a smart contract, it would be convenient for me to see an error like 'You are not admin', /path/to/lib.rs:1:2 instead of panicked at /path/to/lib.rs:1:2:\nYou are not admin. The user will see an error at the beginning of the line instead of long path to source at the beginning.
2. Different versions of rust have different default panic message formats. If you rely on this format in tests, then you will have to change the tests. This also applies to our pallet tests.

[breathx]

🥳