This is a simplified alternative to PR #93, with the same basic idea:
The original implementation creates patches from the diff between subsequent gzipped archives (.tar.gz).
This leads to excessive patch sizes, because small changes in the source can lead to large changes in the gzip-compressed archives. As a result, patches become practically useless.
The new solution adopted in this PR is to create patches from the uncompressed .tar files. This leads to much smaller diffs, making patches useful again.
Note that the archives are still compressed using gzip, in order to save bandwidth and storage space, but they are decompressed before patching, and re-compressed after patching.
One drawback is that patch creation, on the repo-side, now takes a lot longer, simply because an uncompressed archive is (much) larger than a compressed one. Also we are more likely to run into memory size limits, see e.g. 1, 2:
bsdiff is quite memory-hungry. It requires max(17n,9n+m)+O(1) bytes of memory, where n is the size of the old file and m is the size of the new file. bspatch requires n+m+O(1) bytes.
This is a simplified alternative to PR #93, with the same basic idea:
The original implementation creates patches from the diff between subsequent gzipped archives (
.tar.gz).This leads to excessive patch sizes, because small changes in the source can lead to large changes in the gzip-compressed archives. As a result, patches become practically useless.
The new solution adopted in this PR is to create patches from the uncompressed
.tarfiles. This leads to much smaller diffs, making patches useful again.Note that the archives are still compressed using gzip, in order to save bandwidth and storage space, but they are decompressed before patching, and re-compressed after patching.
One drawback is that patch creation, on the repo-side, now takes a lot longer, simply because an uncompressed archive is (much) larger than a compressed one. Also we are more likely to run into memory size limits, see e.g. 1, 2:
fixes #69