diff options
author | Joey Hess <joeyh@joeyh.name> | 2017-10-16 14:10:03 -0400 |
---|---|---|
committer | Joey Hess <joeyh@joeyh.name> | 2017-10-16 14:10:20 -0400 |
commit | 95697161487c5df3d8a88cdccbc40fd0c4d7b094 (patch) | |
tree | 7fb5769ee098ddafb445e587f550a22c4f1f49c9 /doc | |
parent | fb899e75cf1ac84f3fd61ea39288811bacddee2c (diff) |
Avoid repeated checking that files passed on the command line exist.
git annex add, git annex lock etc make multiple seek passes,
and each seek pass checked that files existed. That was unncessary
redundant work.
Fixed by adding a new WorkTreeItem type, make seek actions use it,
and check that the files exist when constructing it.
This commit was supported by the NSF-funded DataLad project.
Diffstat (limited to 'doc')
2 files changed, 35 insertions, 0 deletions
diff --git a/doc/bugs/add_-J_fails_with_not_found/comment_4_983aa78a20672e4d2b1b74a922eeba0a._comment b/doc/bugs/add_-J_fails_with_not_found/comment_4_983aa78a20672e4d2b1b74a922eeba0a._comment new file mode 100644 index 000000000..f997b7193 --- /dev/null +++ b/doc/bugs/add_-J_fails_with_not_found/comment_4_983aa78a20672e4d2b1b74a922eeba0a._comment @@ -0,0 +1,22 @@ +[[!comment format=mdwn + username="joey" + subject="""comment 4""" + date="2017-10-16T16:58:46Z" + content=""" +I was worried there could be further races in the seeking +done by withFilesOldUnlocked and withFilesMaybeModified if those +run while files are still being ingested by actions run earlier +in the `git annex add`. Seems this is not a problem though -- + +withFilesOldUnlocked looks for typeChanged files, but the files +that were just/are currently being added were not in git before, +so are not typeChanged. + +withFilesMaybeModified looks for modified files, and again these +files were/are just being added for the first time, so it won't stumble +over them. + +So, I don't think a synchronization point is needed. In fact, +all three seeks could actually be run more concurrently than they are not +without stepping on one-another's toes. +"""]] diff --git a/doc/bugs/add_-J_fails_with_not_found/comment_5_c18b38456d58900f5710a311eced9f34._comment b/doc/bugs/add_-J_fails_with_not_found/comment_5_c18b38456d58900f5710a311eced9f34._comment new file mode 100644 index 000000000..352cc7a26 --- /dev/null +++ b/doc/bugs/add_-J_fails_with_not_found/comment_5_c18b38456d58900f5710a311eced9f34._comment @@ -0,0 +1,13 @@ +[[!comment format=mdwn + username="joey" + subject="""comment 5""" + date="2017-10-16T17:06:43Z" + content=""" +That leaves only the innefficiency of checkFileOrDirectoryExists being +run three times per parameter passed to `git annex add`. + +There are some other commands that also run checkFileOrDirectoryExists +multiple times. `git annex lock` being one. + +So, I factored that out into a separate pass, that's only done once. +"""]] |