If you can compress an encrypted file, it means you have used a terrible cipher. However, Kolmogorov complexity of such file can be tiny: e.g. of the original file plus of the cipher with crypto key applied.

Also interesting is the other perspective: compression removes redundancy, which could be used for cryptanalysis. Like a cipher, a perfect compressor would produce a looking completely random bit sequence, strongly mixing the information. Hence, a frequent task: compression + encryption could use a weaker requirements for the cipher, e.g. 4 AES rounds instead of 8, saving lots of resources.

Interesting and important topic is combining both tasks into one phase, what doesn’t seem far for strong compressors: put encryption into compression, what can be especially significant for energy restricted devices like Internet of Things, medical implants, RFIDs, smartwatches. Various techniques were proposed for this purpose: using LZ, BWT, Huffman, arithmetic coding ([16] — [22] here). Universal way is putting it into entropy coding and tANS (e.g. LZFSE, ZSTD, lzturbo) is a perfect new tool for this purpose, encryption can be added at nearly no additional cost (by perturbing symbol spread using a (CS)PRNG initialized with crypto key). Huffman can be seen as its degenerated case (symbols spread in ranges of 2^k size), tANS has much more complex behavior (e.g. 3 sources of chaos), the symbol doesn’t longer determine the bit sequence … not even the length of produced bit sequence (floor or ceiling of lg(1/p)).