After Paul's fixes for the lzh-then-lzw bug, I wanted a test case that does not depend on having tools to create lzh or lzw (.Z) files, and I didn't want to check in large opaque .Z or .lzh blobs. So I got a reproducer for the decompress-to-invalid-result case and minimized it.
While doing that, I realized I could likely find a lzh-lzh pair, too. But surprise! when I found one, it exposed that the c_table buffer could leak state, too. So it needed to be cleared just as the other buffers were. Here are three commits I've just pushed:
lzh-lzh-abuse.patch
Description: Binary data
