mirrors/gotosocial
1
0
Fork 0
mirror of https://github.com/superseriousbusiness/gotosocial.git synced 2025-10-31 06:12:25 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

[chore] bump dependencies (#4406)

Browse source 
- codeberg.org/gruf/go-ffmpreg: v0.6.9 -> v0.6.10
- github.com/ncruces/go-sqlite3: v0.27.1 -> v0.28.0
- github.com/stretchr/testify: v1.10.0 -> v1.11.1
- github.com/tdewolff/minify/v2 v2.23.11 -> v2.24.2
- go.opentelemetry.io/otel{,/*}: v1.37.0 -> v1.38.0
- go.opentelemetry.io/contrib/*: v0.62.0 -> v0.63.0

Reviewed-on: https://codeberg.org/superseriousbusiness/gotosocial/pulls/4406
Co-authored-by: kim <grufwub@gmail.com>
Co-committed-by: kim <grufwub@gmail.com>
This commit is contained in:
kim 2025-09-04 15:29:27 +02:00 • committed by kim
commit 78defcd916
274 changed files with 9213 additions and 2368 deletions
Download patch file Download diff file Expand all files Collapse all files

26
vendor/google.golang.org/protobuf/encoding/protowire/wire.go generated vendored
View file

@ -371,7 +371,31 @@ func ConsumeVarint(b []byte) (v uint64, n int) {
func SizeVarint(v uint64) int {
// This computes 1 + (bits.Len64(v)-1)/7.
// 9/64 is a good enough approximation of 1/7
return int(9*uint32(bits.Len64(v))+64) / 64
//
// The Go compiler can translate the bits.LeadingZeros64 call into the LZCNT
// instruction, which is very fast on CPUs from the last few years. The
// specific way of expressing the calculation matches C++ Protobuf, see
// https://godbolt.org/z/4P3h53oM4 for the C++ code and how gcc/clang
// optimize that function for GOAMD64=v1 and GOAMD64=v3 (-march=haswell).
// By OR'ing v with 1, we guarantee that v is never 0, without changing the
// result of SizeVarint. LZCNT is not defined for 0, meaning the compiler
// needs to add extra instructions to handle that case.
//
// The Go compiler currently (go1.24.4) does not make use of this knowledge.
// This opportunity (removing the XOR instruction, which handles the 0 case)
// results in a small (1%) performance win across CPU architectures.
//
// Independently of avoiding the 0 case, we need the v |= 1 line because
// it allows the Go compiler to eliminate an extra XCHGL barrier.
v |= 1
// It would be clearer to write log2value := 63 - uint32(...), but
// writing uint32(...) ^ 63 is much more efficient (-14% ARM, -20% Intel).
// Proof of identity for our value range [0..63]:
// https://go.dev/play/p/Pdn9hEWYakX
log2value := uint32(bits.LeadingZeros64(v)) ^ 63
return int((log2value*9 + (64 + 9)) / 64)
}
// AppendFixed32 appends v to b as a little-endian uint32.