Re: [Cs-studio-core] Repository screwup

[Carcassi, G. <car...@bn...>]

As a general rule: making sure things compile and build is necessary,
but it's not _sufficient_. You can corrupt the repository without
corrupting the build. For example: two people set their editor to
automatically reformat sources... in two different way! Or someone
comments in and comments out portions of files, because he is testing
something, forgets he's doing that, and commits. What gets me every
time, is that in NetBeans there is no way to run a single test, so I
comment out the @Test annotation on the other tests... and then I forget
to change it back.

Never copy working directories, especially across OSes. Depending on the
mechanism, this may change permissions, EOLs, file encoding... Renaming
is fine.

It is not a problem to use the same working directory with different hg
clients: I personally use command line, NetBeans, Eclipse, tortoise...
But whatever you use, when you commit, check the diff. And if there is
something fishy (like Eclipse tells you the file is changes, but then
you look and there is no diff) DON'T commit until you understand exactly
what's going on. DON'T commit because, in the end, it's building.

The repository is now fixed.

Thanks,
Gabriele

-----Original Message-----
From: Knerr, Bastian [mailto:bas...@de...] 
Sent: Monday, August 16, 2010 4:21 AM
To: Carcassi, Gabriele
Cc: CSS-Core
Subject: Re: [Cs-studio-core] Repository screwup

Hi all,
just read the screwup with the line endings - I haven't been here on 
Thursday and Friday, so sorry one for my late response.
Damn, I'm deeply sorry for the trouble on the linux machines, but I just

didn't notice anything weird with the line endings in my pushes.
The tests and the builds did not seem corrupted in any way here at DESY,

we do not build on Linux (perhaps a good idea though).

As to my push routine - I do run the all the tests on all sources (the 
ones we got to work at DESY anyway) and I do check hg out any time 
before a push, but I suppose that this very large changeset for the 
library plugins guava, xmlunit, easymock etc. seemed to have hidden for 
me the load of unrelated file changes due to line endings.
That's my fault for not looking with scrutiny esp. when pushing a lot of

stuff.

First of all I'll check how I introduced the error (I'm using cygwin 
command line for hg and alternately the mercurial eclipse plugin, 
depends on what I do, probably a bad idea), so that we can ensure that 
it doesn't happen again.


To solve the repository screwup:

Isn't it possible to change all the line endings once and for all to a 
format any system can live with and push this large changeset?
In case that's possible, the history wouldn't be destroyed, would it? 
Just a single push of a then very large changeset (for all files) were 
added to the history. So to figure out who is really responsible for a 
line change, you don't pay attention to that specific push.

Best regards,
Bastian



Carcassi, Gabriele wrote:
> Hi,
> 
>  
> 
> I do not know how much people realize things at this point, so let me 
> clarify the situation.
> 
>  
> 
> 1. The following changeset
> 
>
http://cs-studio.hg.sourceforge.net/hgweb/cs-studio/cs-studio/rev/a59465
44e886
> 
> Touched an extensive number of files, changing  line ends (from Linux
to 
> Windows) on many files, including all of the nsls2 projects, and a 
> number of opi projects
> 
>  
> 
> 2. This has broken our build and has essentially made the history 
> useless, as it is now difficult to track which line of code is owned
by 
> whom, when it was committed and why.
> 
>  
> 
> 3. We have been wasting the good part of 2 days to figure out how to 
> revert the situation.
> 
>  
> 
> So, first we need to fix the problem. We have a couple of options.
> 
>  
> 
> 1. We can freeze development, commit everything to the repository. We 
> can do some surgery and remove that commit. Then everybody restarts
from 
> a new cloned repository. This would fix the history. We would perform 
> this tomorrow, at 9 am BNL time.
> 
>  
> 
> 2. We can leave it, and every project is responsible to fix their 
> project files.
> 
>  
> 
> Our preference is number 1. Please, DESY and SNS let me know your 
> preferences as soon as possible, as more time passes the more it
becomes 
> problematic.
> 
>  
> 
> Then we should discuss how to make these problems _/not/_ happen in
the 
> future.
> 
>  
> 
> Thanks,
> 
> Gabriele
> 
> 
>
------------------------------------------------------------------------
> 
>
------------------------------------------------------------------------
------
> This SF.net email is sponsored by 
> 
> Make an app they can't live without
> Enter the BlackBerry Developer Challenge
> http://p.sf.net/sfu/RIM-dev2dev 
> 
> 
>
------------------------------------------------------------------------
> 
> _______________________________________________
> Cs-studio-core mailing list
> Cs-...@li...
> https://lists.sourceforge.net/lists/listinfo/cs-studio-core

-- 
------------------------------------------------------------------------
Dr. Bastian Knerr                        Cryogenic Controls Group(MKS-2)
phone:  +49-40-8998-3883         Deutsches Elektronen Synchrotron (DESY)
fax:    +49-40-8994-4388                                    Notkestr. 85
e-mail: bas...@de...                              22607 Hamburg
web:    http://www-mks2.desy.de/                                 Germany
------------------------------------------------------------------------