Showing code contents in ic-ref causes random test failures

[osa1]

Example from the test closure-params (test/run-drun/ok/closure-params.ic-ref-run.ok):

→ update create_canister()
← completed: (record {1313628723 = service "cvccv-qqaaq-aaaaa-aaaaa-c"})
→ update install_code(record {4849238 = blob ""; 265441191 = blob "\00asm\01\00\00\00\01\a0\81\80\80\00\1a`\n\7f\7f\7f\7f…
...

The blob part in install_code is the wasm module contents. Showing this part causes random breakage as a result of changes in the RTS or elsewhere. One example of this can be observed in #1750 where the test fails with this diff:

-→ update install_code(record {4849238 = blob ""; 265441191 = blob "\00asm\01\00\00\00\01\a0\81\80\80\00\1a`\n\7f\7f\7f\7f…
+→ update install_code(record {4849238 = blob ""; 265441191 = blob "\00asm\01\00\00\00\01\b1\81\80\80\00\1c`\n\7f\7f\7f\7f…

Basically length of the type vector in the type section changes from 32 to 49, which certainly shouldn't break the test.

It'd be great if we could fix this without any more perl lines and wrappers. I suggest we update ic-ref to not print blob contents. Printing some prefix of the type section in wasm modules is probably not useful anyway.

[nomeata]

I’ll fix this. Printing blob prefixes is still useful, I think, but I’ll just recognize asm\0 and handle that.

[nomeata]

Oh, darn, just extending https://github.com/dfinity-lab/ic-ref/blob/master/impl/src/ic-ref-run.hs#L78 doesn’t help; this is a blob nested inside a Candid value. And patching the Candid pretty printer seems wrong…

I’ll just make it print less characters, that should help