Template
1
0
Fork 0
mirror of https://codeberg.org/forgejo/forgejo synced 2024-11-27 20:26:09 +01:00
forgejo/services/pull/lfs.go
zeripath 8598356df1
Refactor and tidy-up the merge/update branch code (#22568)
The merge and update branch code was previously a little tangled and had
some very long functions. The functions were not very clear in their
reasoning and there were deficiencies in their logging and at least one
bug in the handling of LFS for update by rebase.

This PR substantially refactors this code and splits things out to into
separate functions. It also attempts to tidy up the calls by wrapping
things in "context"s. There are also attempts to improve logging when
there are errors.

Signed-off-by: Andrew Thornton <art27@cantab.net>

---------

Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
Co-authored-by: delvh <dev.lh@web.de>
2023-03-07 15:07:35 -05:00

138 lines
4.6 KiB
Go

// Copyright 2019 The Gitea Authors.
// All rights reserved.
// SPDX-License-Identifier: MIT
package pull
import (
"bufio"
"context"
"io"
"strconv"
"sync"
"code.gitea.io/gitea/models/db"
git_model "code.gitea.io/gitea/models/git"
issues_model "code.gitea.io/gitea/models/issues"
"code.gitea.io/gitea/modules/git/pipeline"
"code.gitea.io/gitea/modules/lfs"
"code.gitea.io/gitea/modules/log"
)
// LFSPush pushes lfs objects referred to in new commits in the head repository from the base repository
func LFSPush(ctx context.Context, tmpBasePath, mergeHeadSHA, mergeBaseSHA string, pr *issues_model.PullRequest) error {
// Now we have to implement git lfs push
// git rev-list --objects --filter=blob:limit=1k HEAD --not base
// pass blob shas in to git cat-file --batch-check (possibly unnecessary)
// ensure only blobs and <=1k size then pass in to git cat-file --batch
// to read each sha and check each as a pointer
// Then if they are lfs -> add them to the baseRepo
revListReader, revListWriter := io.Pipe()
shasToCheckReader, shasToCheckWriter := io.Pipe()
catFileCheckReader, catFileCheckWriter := io.Pipe()
shasToBatchReader, shasToBatchWriter := io.Pipe()
catFileBatchReader, catFileBatchWriter := io.Pipe()
errChan := make(chan error, 1)
wg := sync.WaitGroup{}
wg.Add(6)
// Create the go-routines in reverse order.
// 6. Take the output of cat-file --batch and check if each file in turn
// to see if they're pointers to files in the LFS store associated with
// the head repo and add them to the base repo if so
go createLFSMetaObjectsFromCatFileBatch(catFileBatchReader, &wg, pr)
// 5. Take the shas of the blobs and batch read them
go pipeline.CatFileBatch(ctx, shasToBatchReader, catFileBatchWriter, &wg, tmpBasePath)
// 4. From the provided objects restrict to blobs <=1k
go pipeline.BlobsLessThan1024FromCatFileBatchCheck(catFileCheckReader, shasToBatchWriter, &wg)
// 3. Run batch-check on the objects retrieved from rev-list
go pipeline.CatFileBatchCheck(ctx, shasToCheckReader, catFileCheckWriter, &wg, tmpBasePath)
// 2. Check each object retrieved rejecting those without names as they will be commits or trees
go pipeline.BlobsFromRevListObjects(revListReader, shasToCheckWriter, &wg)
// 1. Run rev-list objects from mergeHead to mergeBase
go pipeline.RevListObjects(ctx, revListWriter, &wg, tmpBasePath, mergeHeadSHA, mergeBaseSHA, errChan)
wg.Wait()
select {
case err, has := <-errChan:
if has {
return err
}
default:
}
return nil
}
func createLFSMetaObjectsFromCatFileBatch(catFileBatchReader *io.PipeReader, wg *sync.WaitGroup, pr *issues_model.PullRequest) {
defer wg.Done()
defer catFileBatchReader.Close()
contentStore := lfs.NewContentStore()
bufferedReader := bufio.NewReader(catFileBatchReader)
buf := make([]byte, 1025)
for {
// File descriptor line: sha
_, err := bufferedReader.ReadString(' ')
if err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
// Throw away the blob
if _, err := bufferedReader.ReadString(' '); err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
sizeStr, err := bufferedReader.ReadString('\n')
if err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
size, err := strconv.Atoi(sizeStr[:len(sizeStr)-1])
if err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
pointerBuf := buf[:size+1]
if _, err := io.ReadFull(bufferedReader, pointerBuf); err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
pointerBuf = pointerBuf[:size]
// Now we need to check if the pointerBuf is an LFS pointer
pointer, _ := lfs.ReadPointerFromBuffer(pointerBuf)
if !pointer.IsValid() {
continue
}
exist, _ := contentStore.Exists(pointer)
if !exist {
continue
}
// Then we need to check that this pointer is in the db
if _, err := git_model.GetLFSMetaObjectByOid(db.DefaultContext, pr.HeadRepoID, pointer.Oid); err != nil {
if err == git_model.ErrLFSObjectNotExist {
log.Warn("During merge of: %d in %-v, there is a pointer to LFS Oid: %s which although present in the LFS store is not associated with the head repo %-v", pr.Index, pr.BaseRepo, pointer.Oid, pr.HeadRepo)
continue
}
_ = catFileBatchReader.CloseWithError(err)
break
}
// OK we have a pointer that is associated with the head repo
// and is actually a file in the LFS
// Therefore it should be associated with the base repo
meta := &git_model.LFSMetaObject{Pointer: pointer}
meta.RepositoryID = pr.BaseRepoID
if _, err := git_model.NewLFSMetaObject(db.DefaultContext, meta); err != nil {
_ = catFileBatchReader.CloseWithError(err)
break
}
}
}