add correct test for 1515 #5364
Conversation
ghost
assigned 
| @@ -394,6 +394,27 @@ file_size() { | |||
| $_STAT "$1" | |||
| } | |||
|
|
|||
| directory_size() { | |||
There was a problem hiding this comment.
Why not just something like du -sb --apparent-size?
There was a problem hiding this comment.
- Didn't know about it.
- MacOS 😢.
Stebalien
added
need/review
test/sharness/lib/test-lib.sh
Outdated
| @@ -409,3 +430,9 @@ convert_tcp_maddr() { | |||
| port_from_maddr() { | |||
| echo $1 | awk -F'/' '{ print $NF }' | |||
| } | |||
|
|
|||
|
|
|||
| fail() { | |||
There was a problem hiding this comment.
There is test_fsh - https://github.com/ipfs/go-ipfs/blob/1054826ac49f66d329d3088143f1a7f2b9930dcc/test/ipfs-test-lib.sh#L22-L27 and it's used all over sharness tests - like test_cmp expected unique_hash || test_fsh cat add_output
When checking to see if GC fully reverses an `ipfs add`, we should check the size of the actual files, not the directory sizes. A bunch of empty directories won't use *that* much space and really shouldn't count against GC. closes #1515 License: MIT Signed-off-by: Steven Allen <[email protected]>
Stebalien
added
RFM
and removed
status/in-progress
Then the process is still irreversible. Accumulating empty dirs doesn't sound sustainable either. While the effect is negligible for embedded machines, this is not necessarily the case for clusters. |
|
It's bounded to 1024 empty directories. Whenever we store an object, we store it at |
When checking to see if GC fully reverses an
ipfs add, we should check the size of the actual files, not the directory sizes. A bunch of empty directories won't use that much space and really shouldn't count against GC.closes #1515
I'm not sure why this has been sitting around for so long but I never seem to have created this PR.