Disallow unreachable code sections

[hugo-dc]

This PR disallows EOF unreachable code sections.

Pending:

• Add new unit tests specific for unreachable code sections

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (b74b9db) 97.90% compared to head (b9ee992) 97.91%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #721   +/-   ##
=======================================
  Coverage   97.90%   97.91%           
=======================================
  Files         110      110           
  Lines       10573    10644   +71     
=======================================
+ Hits        10352    10422   +70     
- Misses        221      222    +1     

Flag	Coverage Δ
blockchaintests	59.92% <0.00%> (-0.32%)	⬇️
statetests	62.22% <85.71%> (+0.05%)	⬆️
statetests-silkpre	25.83% <0.00%> (-0.17%)	⬇️
unittests	95.88% <96.93%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
lib/evmone/eof.hpp	100.00% <ø> (ø)
test/unittests/eof_validation_test.cpp	100.00% <100.00%> (ø)
test/unittests/eof_test.cpp	94.11% <85.71%> (+0.78%)	⬆️
lib/evmone/eof.cpp	83.75% <81.81%> (+0.10%)	⬆️

[gumb0]

Please rebase and take into account sections targeted by JUMPF.

[gumb0]

Also please add a validation unit test with unreachable sections error.

[gumb0]

Note also you can now build an EOF bytecode with multiple section using bytecode.hpp utility, like here:

evmone/test/unittests/evm_eof_function_test.cpp

Lines 97 to 100 in 3d0242b

	const bytecode code =
	eof_bytecode(1023 * push(1) + OP_CALLF + "0001" + 1021 * OP_POP + OP_RETURN, 1023)
	.code(push(1) + OP_CALLF + "0002" + OP_POP + OP_RETF, 0, 0, 1)
	.code(push(1) + OP_POP + OP_RETF, 0, 0, 1);

[hugo-dc]

Note also you can now build an EOF bytecode with multiple section using bytecode.hpp utility, like here:

...

In new tests I'm using that approach:

evmone/test/unittests/eof_validation_test.cpp

Lines 1202 to 1204 in 7f1487c

	const auto code = eof_bytecode(bytecode{OP_CALLF} + "0001" + OP_STOP, 0)
	.code(bytecode{"5B"} + OP_RETF, 0, 0, 0)
	.code(bytecode{"FE"}, 0, 0x80, 0);

However, in some other parts, I'm generating the code "manually" in order to generate many code sections where each section calls the next one:

evmone/test/unittests/eof_validation_test.cpp

Lines 1278 to 1281 in 7f1487c

	EXPECT_EQ(
	validate_eof(from_spaced_hex("EF00 01 010400 020100" + section_size_256 + "040000 00" +
	section_types_256 + code_sections_256_err_001)
	.value()),

Not sure if there is a way to do something like that by using the bytecode.hpp utility.

[gumb0]

I found a more concise way of making these sequences, I think I'll make a separate PR with it.

[gumb0]

I did it in #796, please rebase now and use similar approach here.

[gumb0]

It would make ssense to use code_sections_1024 in invalid case, too, just adding one more after them.

[hugo-dc]

I added another variable code_sections_1025 in order that section 1024 is calling section 1025 (so we don't have unreachable code sections).

[chfast]

Issues I reported were fixed.

[gumb0]

Not sure this does what you intended, maybe loop index is wrong, e.g. code_sections_256_err_254 seems to end up with a sequence jumpf{1} jumpf{2} jumpf{1} jumpf{2} jumpf{3} ...
I think intention was to have jumpf{1} jumpf{2} jumpf{3} jumpf{4} ...

Also, if you want to use bytecode here, it should be possible with something like:

auto code_sections_256_err_001 = eof_bytecode(bytecode{JUMPF} + "0001").code(bytecode{JUMPF} + "0002", 0, 0x80, 0);
for (int i = 2; i < 254; ++i)
  code_sections_256_err_001.code(bytecode{JUMPF} + big_endian(static_cast<uint16_t>(i + 1)), 0, 0x80, 0); 

(Also would be nice to have helpers in bytecode.hpp for callf / jumpf to not write bytecode{JUMPF} every time, but that's probably better for separate PR)

[hugo-dc]

There was an error in the loop indices, it is fixed and it is now using bytecode.

[gumb0]

Looks fine to me now, could you squash the commits?

[hugo-dc]

Looks fine to me now, could you squash the commits?

Commits squashed