feat: Support DWARF debug info in PE files

[casept]

I'm not quite sure how I should implement tests for this. All the other formats seem to use breakpad binaries as the test case, but breakpad_client doesn't build under MinGW. Is it relevant that the test cases match? Or can I just use some random project for this?

Also, is support for platforms with non-4K page size relevant? Goblin doesn't seem to expose page size, so this metadata would have to be kept up-to-date in symbolic.

Closes #352.

[codecov]

Codecov Report

Merging #744 (2a51d5b) into master (bad5cb8) will increase coverage by 0.37%.
The diff coverage is 95.55%.

❗ Current head 2a51d5b differs from pull request most recent head a032761. Consider uploading reports for the commit a032761 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #744      +/-   ##
==========================================
+ Coverage   73.43%   73.80%   +0.37%     
==========================================
  Files          69       69              
  Lines       14875    14875              
==========================================
+ Hits        10923    10979      +56     
+ Misses       3952     3896      -56     

[casept]

I've snapshotted a binary of the sqlite3 shell I compiled for the test cases now.

[loewenheim]

Looks good to me, thank you very much for this!

[Swatinem]

Just a NOTE: removing these types is considered a breaking change.

As getsentry/publish#1700 has been stuck and noone has tried to fix this yet, we can squeeze this breaking change into the next release still.

Please make sure to document the changes in the changelog though.

[casept]

I've updated the changelog.

[Swatinem]

Also, 3+M for the testcase is a bit hefty. All you would need is a main fn with a printf, just enough to get some kind of debug info out of it.

[casept]

Also, 3+M for the testcase is a bit hefty. All you would need is a main fn with a printf, just enough to get some kind of debug info out of it.

The reason why I didn't do this is because the other test fixtures looked more like "real world" programs. The MacOS example is also over 1M in size.

Anyways, the test cases are replaced now.

[Swatinem]

Clippy still complains about unneeded return statements. After that I would say its good to go :-)

[casept]

Clippy still complains about unneeded return statements. After that I would say its good to go :-)

Fixed!