apt/apt-pkg/pkgcache.cc, branch 2.5.2

Invalidate cached architecture list when building cache

2021-10-19T16:25:51Z

Fix a regression in python-apt where switching the architectures in the config between cache invocations regressed. Regression-Of: 8ff4e226af55a9feb168477a2b1a99f9c5152e54 Gbp-Dch: full

All pkgCaches are MultiArch caches

2021-09-04T13:35:15Z

Back in 2015 the code inside libapt who was using this field was dropped as even if we are on a system which is not configured for MultiArch, there are still edge cases in which the cache can include very foreign packages, so any assumption you could make thinking only a single architecture will be in the cache is probably wrong. Maintaining two different codepaths for Multi- and SingleArch is likely not very beneficial for code and users alike and is surprisingly hard to answer correctly and becoming even harder still, so always assuming the "worst case" seems like the far better option. References: 6c9937da76b9155d166092b9dda22d06200510c1

Free XXH3 state to avoid leak in cache hashing

2021-02-03T16:36:45Z

We do this once (usually), so the leak is tremendously big, but it is detected as a leak by the fuzzer and trips it up.

Unroll pkgCache::sHash 8 time, break up dependency

2020-12-15T21:13:12Z

Unroll pkgCache::sHash 8 times and break up the dependency between the iterations by expanding the calculation H(n) = 33 * H(n-1) + c 8 times rather than performing it 8 times. This seems to yield about a 0.4% performance improvement. I tried unrolling 4 and 2 bytes as well, those only having 3 ifs at the end rather than 1 small loop; but that was actually slower - potentially the code got to large and the cache went bonkers. I also tried unrolling 4 times instead of 8, thinking that smaller code might yield better results overall then, but that was slower as well.

Use XXH3 for cache, hash table hashing

2020-12-15T12:47:22Z

XXH3 is faster than both our CRC32c implementation as well as DJB hash for hash table hashing, so meh, let's switch to it.

Raise APT::Cache-HashtableSize to 196613

2020-12-10T14:40:08Z

We now have over 100k package names, my Ubuntu system has 125k arleady, so increase the hash table size to match, this will cost us about a MB in cache size, but give a very nice speed up somewhere around 3%-4% or so.

Allow FMV SSE4.2 detection to succeed on clang

2020-05-25T10:05:00Z

As the builtins were used in the feature test also in the default branch clang fails to compile the test helpfully complaining that you need to compile with sse4.2 to use that while on gcc it is optimized out as unused code and produces only a warning for that… removing the code from the default branch fixes this problem, but we adapt the code some more to avoid compilers optimizing it out in the future just in case.

Make map_pointer typesafe

2020-02-24T17:29:07Z

Instead of just using uint32_t, which would allow you to assign e.g. a map_pointer to a map_pointer, use our own smarter struct that has strict type checking. We allow creating a map_pointer from a nullptr, and we allow comparing map_pointer to nullptr, which also deals with comparisons against 0 which are often used, as 0 will be implictly converted to nullptr.

pkgcache.cc: Mix PACKAGE_VERSION into the cache hash

2020-01-16T17:21:20Z

This ensures that caches build with one version can't be opened with another, which makes sense. It's a temporary approach until we can replace major:minor fields with a version string. For example, this would have prevented 1.9.7 from using broken caches from 1.9.6.

Rename _count() macro to APT_ARRAY_SIZE()

2020-01-07T21:51:41Z