archive: switch offsets and lengths to unsigned 32-bit integers

[rokm]

Switch the offsets and lengths in CTOC and CArchive python data structures from signed 32-bit integers to unsigned 32-bit integers. Adjust the field definitions in bootloader's CArchive reader code accordingly.

Increases the maximum allowed CArchive size from 2 GiB to 4 GiB.

Fixes #3939.

[rokm]

This is a "lite" branch, which does not include tests. Those were useful for development and local testing, but on the CI they would need to remain disabled anyway (due to storage requirements, which could be further exacerbated due to potential timeouts).

The full branch is here, if anyone's interested: https://github.com/rokm/pyinstaller/tree/4gb-archive.
(The large-data tests are disabled by default, and need to be explicitly enabled on command-line).

[bwoodsend]

I'm guessing that, during the period between merging this and our next rebuilding the bootloaders, we aren't going to experience issues due to mismatch between Python using uint32 and C still using int32 simply because the two types look identical provided you stay in the range of 0 to 2³¹?

[rokm]

I'm guessing that, during the period between merging this and our next rebuilding the bootloaders, we aren't going to experience issues due to mismatch between Python using uint32 and C still using int32 simply because the two types look identical provided you stay in the range of 0 to 231?

Until bootloaders are rebuilt, the difference in behavior will be that the python code will happily pack the large CArchive, but the old bootloader will misinterpret the large offsets/sizes as negative. Which means that the error will change from being built-time to being run-time. If that's unacceptable, we can delay merging this until we're ready for new release.

EDIT: for CArchives smaller than 2 GiB, the behavior should be unchanged, as the change in representation involves only the most-significant bit.

[bwoodsend]

EDIT: for CArchives smaller than 2 GiB, the behavior should be unchanged, as the change in representation involves only the most-significant bit.

That's what I wanted to hear.