timestamps not git-cloned

Gentlemen, it's my first git-clone, $ git-clone git://git.debian.org/git/pkg-fso/files.git and I'm disappointed to find the timestamps of the files created are all now and not the date of last edit. At least mention something about this on the git-clone man page.

Hi,

On Fri, Nov 28, 2008 at 7:54 AM,  <jidanni@jidanni.org> wrote:

I do not think there is an VCS that records timestamps. Only file
contents are tracked. Also, if you clone from systems across time
zones, what time do you expect to set on the files. IMO, it is not
practical to track timestamps.
 Are you concerned of 'make' doing a complete build when you switch
branches? I guess it makes sense only in that scenario. I wish 'make'
had some feature to track changes instead of timestamps alone...

-dhruva

d> Also, if you clone from systems across time zones, what time do you d> expect to set on the files.

I'm just used to tar, cpio, scp -a, rsync -a, ar, etc. using 'date +%s' seconds internally, so no timezone problem.

I hate it when I get some latest WhizBang.tgz, only to untar it to find all the files' dates the same, when in fact the README hasn't been touched in seven years, but you can't tell that from ls -l. I recall some content tracker was involved.

Of course I'll allowing you to know my delicate first day impressions. I'm sure as I grow older I will learn the difference between content tracker and archiver.

On Fri, 28 Nov 2008, jidanni@jidanni.org wrote:

Well, README was just touched; it wasn't on your disk at all shortly before. This would make a big difference if, for example, you unpacked "foo-1.0" on top of "foo-1.1" and the timestamps were from when the files were originally created, and now all of the source files that changed are older than the object files and the build system does nothing.

Of course, with archives, you don't unpack different versions into the same directory, but with a version control system, you'll do it all the time, so you really need the system to put on disk the times when those files were last put there. If you want to know when the README you've got is from (and a whole lot more) "git log README" will tell you, although it won't tell you if somebody yesterday changed the README they're distributing from some other text to a file that's been sitting on their disk untouched for seven years.

	-Daniel
*This .sig left intentionally blank*

On Fri, Nov 28, 2008 at 01:06:00PM +0800, jidanni@jidanni.org wrote:

If this is the important bit, perhaps git-archive could be changed to create tarballs with file timestamps based on their commit dates.

- Chris

Chris Frey wrote:

Based on the principle of least surprise, I'd consider this a rather good idea.

Stephen R. van den Berg wrote:

Unless I'm missing something, this would make git-archive rather more expensive than it is now: Tree objects do not record any timestamps, so figuring out the last commit that changed a file requires a full history walk in the worst case[*]. (This is another side-effect of not versioning files.) On the other hand, current git-archive's running time depends only on the size of the tree-ish given, including all subtrees and blobs.

My unscientific guesstimates on how much work this would be, in a random (old) linux-2.6 clone:

  $ git rev-parse HEAD
  e013e13bf605b9e6b702adffbe2853cfc60e7806
  $ time git ls-tree -r -t $(git rev-list HEAD~5000..HEAD) >/dev/null

  real    0m1.385s
  user    0m1.164s
  sys     0m0.220s
  $ git rev-list HEAD | wc -l
  117812

So reading (and dumping) all those trees and subtrees incurs a penalty on the order of 30 seconds. Compare to the current running time of git-archive:

  $ time git archive --format=tar HEAD >/dev/null

  real    0m2.790s
  user    0m2.684s
  sys     0m0.072s

Of course, the ratio will keep getting worse as history gets longer.

- Thomas

[*] I think to really have a "worst case" here, you need at least one file in every leaf directory that has not changed since the root commit, and another that changes in every commit to force the search to really read every subtree.

Well all I know is from the simple user who does e.g., # aptitude install linux-doc-2.6.26 # ls -lt /usr/share/doc/linux-doc-2.6.26/Documentation/ he thinks "gosh, can't tell what's new vs. what hasn't changed in years".

OK, now I know why this is tolerable upstream: they all use git.

But for the lowly user downstream who gets what git-archive produces, it seems like a step backwards: "who threw away the timestamp of when each file was last changed?".

OK, http://git.or.cz/gitwiki/ContentLimitations says this is by design.

And OK, thinking "file by file" is old fashioned, I read. The non-git end user should just get used to reading ChangeLogs, if any, and stop doing ls -lt.

But you must admit, /usr/share/doc/linux-doc-2.6.26/Documentation/ etc. are aimed for reading without git.

Anyways, if just in case any individual file modification time information can still be pried from the 40 byte IDs or whatever, I would suggest using it by default in git-archive at least, and maybe even git-clone etc.

Just letting you know my 'valuable first impressions'. I expect once I start smoking more of this "git" stuff, I too will become comfortably numb to aforementioned lowly user problem, so you would never know unless I hereby first told you before it was too late.

jidanni@jidanni.org wrote:

They won't be able to do that anyway, since a spelling correction would update the timestamp anyway. The only way of finding out if there are *content* changes (which is really what matters) is to use some sort of history-browser for those documents. Git is good for that. The sort of people who really need to know when the docs change can be exptected to have a higher technical knowledge than the end-users, so it's not too much to ask that they use such a history browsing tool to find out what's new (or simply "diff").

Either way, timestamps on documents are a very poor way of finding out when something really changed.

Well, /usr/bin/less doesn't require git to function, so I fail to see what the fuzz is all about.

It can, but it's a fairly expensive operation, tracking each files SHA1 backwards in time to see when the commit was done that last made any changes to it. It's not something you want to do for an archive containing 26K files. Trust me on this.

I see a lot of ranting but no patches from you. Since you're the one with the itch, why not just submit a patch and see if distro packagers start using it?

Some words of advice though; Make it optional, or it'll be dropped on the floor. For bonus points, make it calculate timestamps only for a path-spec delimited set of files. That'll cut down expected runtime by a *huge* amount for something like the linux kernel.

Andreas Ericsson <ae@op5.se> writes:

I think the old idea of 'tree blame' (search archives) would be a good interface for this feature... if you decide to write it.

On Sat, Nov 29, 2008 at 3:46 PM, Thomas Rast <trast@student.ethz.ch> wrote:

How many people use git-archive and how many times a day do they use it? For example, kernel.org seems to put out linux-2.x.y.z.tar.bz2 once every 2 to 7 days.

The overhead of this new option (and certainly it should be an option, not the default) should be measured not against the old running time, but against the frequency of usage of the tool. Look at it on those time scales, it may not be a big deal.

By all accounts, this overhead will not affect the "giterate" [meaning git-literate ;-)] people too much.

On Fri, Nov 28, 2008 at 08:38:06AM +0530, dhruva wrote:

Perforce does, at least optionally. But, it's model is so different, I'm not sure it really applies here. I believe it can either be an individual file setting, or for a whole workspace.

David

Hi!

dhruva:

CVS sort of does. The initial checkout sets the time stamp of files to their last *check-in* time (and you can import a file with -D to set its commit time to the current timestamp). Updates do, however, set to current time (to not break make and friends).

I miss this behaviour in Git, but I have learnt to live with it. I guess it is like a difference in philosophy on what time-stamps are supposed to record, like how UNIX "cp" sets the time of the new-born copy to now, while DOS "copy" sets it to the old time-stamp. Coming to Unix and Linux from DOS (via OS/2), I find "cp" behaviour weird. But have learnt to live with it (and use "cp -a" a lot).

Hi,

On Fri, 28 Nov 2008, jidanni@jidanni.org wrote:

You are mistaken to be disappointed. Granted, to a hammer, everything looks like a nail. But it might make more sense to be gentle, and insert the syringe by hand. Maybe it will get less painful that way, too.

In other words, do not be surprised when a source code management tool turns out to be lousy at recreating meta-data -- which has as much to do with source code than a syringe with a nail.

Hth, Dscho

fredag 28 november 2008 03:24:04 skrev jidanni@jidanni.org:

I recommend every new Git user to scan the FAQ. It's not just clone, it's in everything git does in the file system. There is a very good reason git behaves this way in general, although clone could be exception, but then we would have a ton of questions about that inconsistency.

-- robin

Hi,

On Fri, 28 Nov 2008, Robin Rosenberg wrote:

No, clone cannot be an exception. The information is just not stored, and for a good reason: a file's timestamp is nothing you want to commit in source code management. It just does not make sense at all.

Ciao, Dscho