| 1 |
Mike Frysinger posted on Fri, 17 Jun 2011 12:44:52 -0400 as excerpted: |
| 2 |
|
| 3 |
> On Friday, June 17, 2011 11:31:43 Duncan wrote: |
| 4 |
>> It's worth pointing out that if Mike and others' workflow already |
| 5 |
>> involves a lot of this, they'd be modifying it very little if they |
| 6 |
>> simply avoided separate removals. In fact, in borderline cases where a |
| 7 |
>> trivial change may not have made it on its own, as it waited for a |
| 8 |
>> bigger change to come along to be worth doing, the removals combined |
| 9 |
>> with the trivial change may now trigger the trivial change commit |
| 10 |
>> earlier than it would have occurred otherwise. |
| 11 |
> |
| 12 |
> if you look at my commit behavior, this is exactly the sort of thing i |
| 13 |
> avoid. |
| 14 |
> my cvs commits are pretty logically clean to the point where importing |
| 15 |
> into git would result in nice behavior. which means i make one commit |
| 16 |
> to remove, one commit to fix a specific bug, one commit to version bump, |
| 17 |
> etc... |
| 18 |
|
| 19 |
Good point and exactly the behavior best on git as it makes for far |
| 20 |
easier and more effective git bisects when necessary. Unfortunately (for |
| 21 |
oh so many reasons!!), Gentoo's main tree and workflow isn't "git-ified" |
| 22 |
yet. But I can certainly commend someone whose personal standards demand |
| 23 |
that same one-thing-and-one-thing-only commit separation, modern dVCS or |
| 24 |
not. |
| 25 |
|
| 26 |
Meanwhile, case-in-point of why changelogging removals matters. My last |
| 27 |
post was to a kde list, helping someone trying to build kdelibs on RHEL. |
| 28 |
He was missing the libdbusmenu-qt dependency, which I was able to point |
| 29 |
out, and I went on to describe the version info. Gentoo's kdelibs-4.6.4 |
| 30 |
dependency for that library is >= libdbusmenu-qt-0.3.2, but I have 0.8.2 |
| 31 |
installed. |
| 32 |
|
| 33 |
Because the information was in the changelog, I was able to tell him that |
| 34 |
my current 0.8.2 was introduced in April, the other available version on |
| 35 |
gentoo, 0.6.2, was introduced in Sept. 2010, there was a version jump (at |
| 36 |
least on gentoo) between 0.3.5 (from June, 2010) and 0.6.2, and the 0.3.2 |
| 37 |
that's gentoo's minimum requirement was introduced on Gentoo in April |
| 38 |
2010 and removed in Sept, 2010. So even 0.3.2 isn't much more than a |
| 39 |
year old (on RHEL 5 it's likely an upgrade!), but was already considered |
| 40 |
old enough to remove ~6 months later. |
| 41 |
|
| 42 |
That information on 0.3.2 removal wouldn't have been available to me (at |
| 43 |
least not without making a huge project of it, checking Gentoo's viewCVS |
| 44 |
logs on the web) had someone not put it in the changelog. Users DO find |
| 45 |
that information useful and there have been quite a number of times I |
| 46 |
personally have found it useful in helping people not necessarily on |
| 47 |
gentoo (tho I believe I've spotted hugely outdated based on changelogs |
| 48 |
versions of packages on gentoo-users systems, too), but in other parts of |
| 49 |
the FLOSS community. |
| 50 |
|
| 51 |
Having that information not available locally on my system, either by |
| 52 |
changelog as now, or by git whatchanged, if users finally get access to |
| 53 |
direct git-pull once the main tree is git-upgraded, would be a serious |
| 54 |
regression. |
| 55 |
|
| 56 |
-- |
| 57 |
Duncan - List replies preferred. No HTML msgs. |
| 58 |
"Every nonfree program has a lord, a master -- |
| 59 |
and if you use the program, he is your master." Richard Stallman |
Archives