Warnings at compile time

[hxim]

Warnings at compile time

[GotthardtZ]

@L3P3
Thanks for the suggestion. I may acually do that. (Or use a 7-byte array as another union there. )
Compiler warnings are better to be eliminated.

[GotthardtZ]

Thank you for your alertness and reporting these warnings!

These are false positives. The solution in ContextMap is not compiler-friendly, I have to admit.

The compiler is confused by the fact that we are naming each of the 7 bytes in HashElementForContextMap independently (so they look like just 7 independent bytes) and at the same time we are transitioning from one to another with the p pointer (so we are treating the HashElementForContextMap as a 7-byte array). The warning is about this transition.

We could make the warning disappear, if we'd have a 7-byte array in HashElementForContextMap, but then we wouldn't be able to clearly reference these state bytes independently (such as bitState and bitState0, bitState1, etc.)

[L3P3]

You could use constants for the names like byte[MAP_BITSTATE0].

[rafael2k]

Knowing these warnings are false positives is good!
So far, with simple binary data, I got no problem at all. I'm using -3, is that ok? I get good results, sometimes better than higher levels.

[GotthardtZ]

Yes, any compression level is good to use. Lower levels use less memory thus it can't remember patterns encounterd too far in the "past". It makes the modelling more adaptive.
99% of the cases higher levels give better compression ratio, but there are some files with radically changing file content where lower leves work better.