pe: align section reads to FileAlignment, pad sections to SectionAlignment

[williballenthin]

closes #435

[atlas0fd00m]

hey @williballenthin ,

can you help me understand here? the unittests seem to indicate pretty significant failures that don't immediately appear to be "more accurate." i could be mistaken.

thanks!
@

[atlas0fd00m]

specifically, this PR is breaking analysis of Firefox.exe in ways that it should not. any idea why?

[williballenthin]

sure, let me dig into the test failures. i'm glad test cases caught a potential regression here.

we're in the realm of "what happens if sizes (with and without alignment) are pretty weird?" so I'm not super confident in the intuition/reasoning behind this PR. so, i'll also do some bulk runs against a moderate number of PE files and see what happens.

[williballenthin]

sidebar issue: i cannot see the CI results (without logging into CircleCI which requires giving access to my private repos, which i object to on principle). maybe another reason to consider GH Actions?

[atlas0fd00m]

ahhh. these failures look much better.
i'd say it's probably safe to go through https://app.circleci.com/pipelines/github/vivisect/vivisect/342/workflows/aa66fd4a-6c8c-4fee-bb83-f49cc9e446e1/jobs/372 and modify the tests to match what we're actually getting now. these failures indicate an increased size, not a loss of analysis/locations.

the increased Undefined answer is just a sad artifact of adding padding to the workspace, and likely doesn't indicate any regression.

make those few changes and the tests should pass. if you do it quickly, i'll merge it in before 4:30pm EDT. otherwise, @rakuy0 will have to do it since i'll be gone.

[williballenthin]

ok no unexpected failures when i run these tests locally. should be good to go.