ii  git-annex                                     8.20210223-2ubuntu2 

according to the document

Once you've told git-annex what files are large, both git annex add and git add/git commit -a will add the large files to the annex and the small files to git.

then I have .gitattributes

*.gz annex.largefiles=(largerthan=0)
*.tar annex.largefiles=(largerthan=0)
*.zip annex.largefiles=(largerthan=0)
*.bin annex.largefiles=(largerthan=50mb)

When I do git annex add *.gz it change all the file into symlink however when I do git add *.gz, I hope it does the same. But instead, it copy all the file into .git/annex but does not create symlink.

What did I do wrong here?

ii  git-annex                                     8.20210223-2ubuntu2 

according to the document

Once you've told git-annex what files are large, both git annex add and git add/git commit -a will add the large files to the annex and the small files to git.

then I have .gitattributes

*.gz annex.largefiles=(largerthan=0)
*.tar annex.largefiles=(largerthan=0)
*.zip annex.largefiles=(largerthan=0)
*.bin annex.largefiles=(largerthan=50mb)

When I do git annex add *.gz it change all the file into symlink however when I do git add *.gz, I hope it does the same. But instead, it copy all the file into .git/annex but does not create symlink.

What did I do wrong here?

• git
• git-annex

2 Answers 2

If the operation created an object under .git/annex/objects/, then the command works as advertised and has added it into the annex.

Files handled by the git-annex git add filter are added as unlocked, as documented under the annex.addunlocked setting. (In part because, as far as I remember, a smudge/clean filter cannot replace the original with a different type in the first place.) See man git-annex unlock.

Instead of a symlink, the smudge/clean filter stores the Git object as a special text file starting with /annex/..., which you can see using git show HEAD:<path>.

looks like this tool is very buggy. Probably not reliable enough to use, here is some tests:

git annex config --get annex.largefiles
anything
#> anything
# so in theory both git add . and git annex add . should use annex
# but apparently no
# first lets do git annex add
git annex add medicines.jpg 
#> add medicines.jpg 
#> ok                                
#> (recording state in git...)
# medicines.jpg -> .git/annex/objects/VP/4F/SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg/SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg

so looks normal, unlock will use reflink copy, we can verify the BLOCK-RANGE.

git annex unlock medicines.jpg
xfs_bmap -v medicines.jpg 
medicines.jpg:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL
   0: [0..23]:         2776..2799        0 (2776..2799)        24 100000
find .git/annex/objects/ -type f -exec xfs_bmap -v {} \;
.git/annex/objects/VP/4F/SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg/SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL
   0: [0..23]:         2776..2799        0 (2776..2799)        24 100000

git add will not do what is advertised, the stat is regular file not symlink.

git add pts02787667.jpg

Changes to be committed:
  (use "git rm --cached <file>..." to unstage)
        new file:   medicines.jpg
        new file:   pts02787667.jpg


stat pts02787667.jpg 
  File: pts02787667.jpg
  Size: 173414          Blocks: 344        IO Block: 4096   regular file
Device: 821h/2081d      Inode: 148         Links: 1

.git/annex/objects
├── gP
│   └── VF
│       └── SHA256E-s173414--bd553b6abb1ab277e9141257c2b47d5c7278cbe05e7963a11254257cbc7b1949.jpg
│           └── SHA256E-s173414--bd553b6abb1ab277e9141257c2b47d5c7278cbe05e7963a11254257cbc7b1949.jpg
└── VP
    └── 4F
        └── SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg
            └── SHA256E-s11172--839e5e989dd90d32252c1035bfa270613b4d02934c5f2db860851dfa5f13019a.jpg

we can verify the file is reflink (CoW):

xfs_bmap -v pts02787667.jpg 
pts02787667.jpg:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL
   0: [0..343]:        2800..3143        0 (2800..3143)       344 100000

.git/annex/objects/gP/VF/SHA256E-s173414--bd553b6abb1ab277e9141257c2b47d5c7278cbe05e7963a11254257cbc7b1949.jpg/SHA256E-s173414--bd553b6abb1ab277e9141257c2b47d5c7278cbe05e7963a11254257cbc7b1949.jpg:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL
   0: [0..343]:        2800..3143        0 (2800..3143)       344 100000