Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Thu, 14 Dec 2017 11:02:55 -0600

Hi Dirk,

If you're likely to be making changes across multiple modules you can
also use the checkout-modules.sh script in the core's .ci directory
which for each submodule will checkout and pull the branch named in the
.gitmodules file for that module. I haven't tried using it when I have
made local commits that haven't been merged yet though, it may prompt
you for a log message in that case.

You are also leading the way with your question about how to do merge
proposals with the new structure. Unfortunately I think we may have made
this much more complicated for some developers, although as long as your
changes only affect one submodule it shouldn't be too bad. You just have
to make sure that you set the right submodule branch that you want to
merge the change into when you're submitting the proposal to Launchpad.

I do want to have a discussion about the process in the new year, we
need to properly document it, which we previously did on the Wiki page
https://wiki-ext.aps.anl.gov/epics/index.php/How_to_check_out_EPICS_Base_with_Bazaar

There are probably many things out of date that are linked from the V4
page http://epics-pvdata.sourceforge.net/development.html

- Andrew

On 12/14/2017 10:21 AM, Michael Davidsaver wrote:
> On 12/14/2017 08:48 AM, Dirk Zimoch wrote:
>> Hi everyone,
>>
>> What I still do not understand is how to work on the EPICS 7 submodules.
>>
>> On 14.12.2017 14:38, Michael Davidsaver wrote:
>>> I avoid the redundant download with:
>>>
>>>> git clone --branch core/master https://github.com/epics-base/epics-base.git
>>>> git submodule update --init --reference .
>>>
>>> Note that the second line ends with a dot '.'.
>>
>> After that, what is the state of the submodule? Latest commit? I found that git tells me that they are "not on any branch". Is that a problem?
> 
> When 'git status' reports no changes in the super-module, then each of the submodules are at the commit associated with the super-module
> commit.  This will often be a detached HEAD (aka. not on any branch).  This is fine if you're just going to build.  If you want
> to do work, then generally the first step is to go into a submodule (eg. modules/ca) and create a branch.  eg.
> 
>> cd modules/ca
>> git checkout -b mybranchname
> 
> This will create a new branch at the present rev without modifying the working tree (it's safe).
> You can then move that branch around as necessary.  eg.
> 
>> git merge --ff-only origin/ca/master
> 
> Which is the safe way to jump to the latest rev.  If will abort if the current branch has diverged,
> or if uncommitted changes would be clobbered.
> 
> 
>> So when I want to propose a change for merge, what should I do?
>> Create a new branch I suppose but in the submodule or in the main directory?
> 
> If you're only changing one submodule, then go into the submodule directory and forget about
> the supermodule while you're working.  To continue my modules/ca example:
> 
>> cd modules/ca
>> git checkout -b mybranchname
>> ... file changes ...
>> git commit -m 'Make CA great again"
>> git remote add md https://github.com/mdavidsaver/epics-base.git
>> git push md mybranchname
> 
>> Then pushing the merge request, will I push from the submodule or from the main directory?
>>
>> The whole work flow with submodules is still very unclear to me.
>>
>> Dirk
>>
>>>
>>>
>>> On 12/14/2017 02:20 AM, Torsten Bögershausen wrote:
>>>> Hej,
>>>> Yesterday I was sitting on a 3Mbit/sec line I was cloning base from Git,
>>>> I realized that "base" needs to be transferred 4 times.
>>>> I got 4 pack files, each 95M.
>>>> One for base, modules/ca, modules/database, modules/libcom/
>>>>
>>>> 99 234 251 ./.git/modules/modules/ca/objects/pack/pack-08cf7dbc8f8108176628acb637cfa5d864f0aa43.pack
>>>> 99234251 ./.git/modules/modules/database/objects/pack/pack-08cf7dbc8f8108176628acb637cfa5d864f0aa43.pack
>>>> 99234251 ./.git/modules/modules/libcom/objects/pack/pack-08cf7dbc8f8108176628acb637cfa5d864f0aa43.pack
>>>> 99234251 ./.git/objects/pack/pack-08cf7dbc8f8108176628acb637cfa5d864f0aa43.pack
>>>>
>>>> I think this is because the modularization was done without running
>>>> "git filter-branch".
>>>>
>>>> My first question:
>>>> Does anybody else feels that this is a problem to download
>>>> 4*95 Mbyte, when 1*95 Mbyte enough ?
>>>>
>>>> And if yes, what should we do about it?
>>>> a) Undo the modularization.  Keep ca, database, libcom in the same
>>>>    Git as before ?
>>>> b) Run Git filter-branch and try a surgery on the Gits?
>>>> c) Leave everything as it is ?
>>>>
>>>> Any thoughts and opinions ?
>>>>
>>>>
>>>>
>>>>
>>>> On 01/12/17 01:20, Andrew Johnson wrote:
>>>>> I just tagged, tarred, tested uploaded and pushed the -rc1, but I don't
>>>>> have time to write the announcement email for tech-talk tonight. If no
>>>>> one else feels like doing so I'll get to it tomorrow morning.
>>>>>
>>>>> - Andrew
>>>>>
>>>
> 

-- 
Arguing for surveillance because you have nothing to hide is no
different than making the claim, "I don't care about freedom of
speech because I have nothing to say." -- Edward Snowdon

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System