HG bundle file size

[Sebastian Unger]

Hi Matt et al,

It took a bit longer than I anticipated, but I have now had a chance to run
the experiment with --type none:














*seb at ws3559:~/Documents/TTDE/TerminalsValidation$ hg bundle -r 'draft()'
--base 'public()' --type none all.hg; for r in $(hg log -r 'draft()'
--template '{node} '); do hg bundle -r $r --base "${r}^" --type none
${r}.hg; done4 changesets found1 changesets found1 changesets found1
changesets found1 changesets
foundseb at ws3559:~/Documents/TTDE/TerminalsValidation$ ls -ltotal
824-rw-r--r-- 1 seb users    624 Aug  6 13:11
3f02e6f97ee3ad12bf1e881a0765787f7334b545.hg-rw-r--r-- 1 seb users   1308
Aug  6 13:11 7efcf424dea3b322889ee0c5f3e527aee003ff10.hg-rw-r--r-- 1 seb
users    921 Aug  6 13:11
aa2b98f4ae401ac09b6ccada16bd5946146b7bfb.hg-rw-r--r-- 1 seb users 824463
Aug  6 13:11 all.hg-rw-r--r-- 1 seb users    741 Aug  6 13:11
d96df5dbf1fd9164f8a8bbeb0982633a15b29e08.hg*

Looking at the files themselves, the all.hg file seems to have a list of
all or most of the files in our repo (just the names, not the contents)
whereas the individual bundles don't. I'm currently in the process of
pursuing the people that could authorize me to disclose these file names
which would allow me to send the bundles themselves if that were helpful.
It might be easier for them to approve me sending this to individuals
rather than an entire list.

Cheers,
Seb


On Wed, Aug 6, 2014 at 8:02 AM, Sebastian Unger <sebunger44 at gmail.com>
wrote:

> Hi Matt,
>
> No, no binary files involved. I'll try the --type none when I get to work.
> But I tried one other thing late last night:
>
> When I bundle up the changesets in question one by one in individual
> files, they are each only a couple hundred bytes long. Maybe a KB
> altogether.
>
> One thing to note about these changesets is, that they're not related.
> They are additional heads added off of historic points in the repo with
> fixes so that these points can be rebuilt using a newer environment. Maybe
> that's a factor here.
>
> I'll let you know what I find with --type none in two hours give or take.
>
> Cheers,
> Seb
> On 6/08/2014 7:29 am, "Matt Mackall" <mpm at selenic.com> wrote:
>
>> On Tue, 2014-08-05 at 22:02 +1200, Sebastian Unger wrote:
>> > Hi there,
>> >
>> > I'm not sure whether this should be directed here or the developer list.
>> > Let me know if I should take this over there.
>> >
>> > I am a bit confused about the size of the files generated with the hg
>> > bundle command. I often use those to transfer or store changesets that
>> I'm
>> > not prepared to push for various reasons. These files seem
>> extraordinarily
>> > large to me. For instance, in the case I'm looking at right now, I'm
>> trying
>> > to bundle up 4 changesets with a couple one-line changes each. The
>> > changesets I want are in draft, everything else is public. For
>> comparison I
>> > also export the same changesets and pass the resulting diff through bz:
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > *seb at eragon:~/Documents/TTDE/TerminalsRoot$ hg bundle -r 'draft()'
>> --base
>> > 'public()' bundled.hg4 changesets
>> > foundseb at eragon:~/Documents/TTDE/TerminalsRoot$ hg export -r 'draft()'
>> |
>> > bzip2 -c >bundled.diff.bz2seb at eragon:~/Documents/TTDE/TerminalsRoot$ ls
>> > -ltotal 208-rw-r--r-- 1 seb users   1700 Aug  5 21:58
>> > bundled.diff.bz2-rw-r--r-- 1 seb users 206878 Aug  5 21:57 bundled.hg*
>> > I thought that an export stores most of the info in a bundle (So far I
>> only
>> > found grafts can't be imported with --exact) and in fact more than a
>> bundle
>> > (diffs have lots of context that is not needed in a bundle).
>> >
>> > Am I missing something here?
>>
>> Yeah, this shouldn't happen.
>>
>> Sure one of those "diffs" doesn't contain a line like "Binary file foo
>> has changed"?
>>
>> If you can't share your bundle with us, you can use 'hg bundle --type
>> none' to get an uncompressed bundle that you can more easily inspect
>> (for instance with 'strings').
>>
>> --
>> Mathematics is the supreme nostalgia of our time.
>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurial-scm.org/pipermail/mercurial/attachments/20140806/af4818ee/attachment-0002.html>