Re: Script Help

[Todd Denniston]

BTW, Your replies are a little better now. Thanks for changing what ever it was.

Rez P wrote, On 02/24/2009 12:33 PM:
> Thanks for taking time and your thorough explanation and examples. 
>  
> I usually do a general or selective update to my 
>  CVS working folders and get only the files that 
>  need to go into a build and tag the CVS module 
>  only if a build is successful.  


So at this point we have *A* tag, that should match cvs_stat.log.


>  Then do a 'cvs -f stat > cvs_stat.log' 
>  which shows my working revisions vs. repo revisions and 
>  I then pack the generated file into the war file as a record.  
>  I was hoping that there was a way to generate a log that 
>  only shows my working copy files and their full path and 
>  their file revisions and I could do away with all the extra 
>  info that cvs -f stat generates.
>  

Do you mean some kind of summary from 'cvs log' here?

>  
> I prefer to tag CVS after doing a build.  
>  What if the build keeps breaking?  

keep making test tags, they are cheap.

>  Then each time until it's fixed, 
>  I have to do more steps tagging or moving the tags to the new 
>  fixed files as they're checked in, 
>  and many times I've forgotten to retag, 
>  therefore, fetching the wrong revs.  

I think the fundamental difference you and I have here is that you do release 
builds from your sandbox.

I only do them from checked out tags[1].  which means I get my sandbox 
working, tag it (test_tag), move the the release build directory(RBD), clean 
the RBD, checkout based on the tag and build. If everything built and worked 
correctly in the RBD, I apply a release build tag (RBT) to the test_tag, and 
truly do a release from the RBT, i.e., I do another clean checkout, but using 
the RBT and release that.  my method makes sure that everything is required 
for the build IS in CVS.


[1] When I am the Configuration Manager, I check out code, never make it. 
Making it all work together is the teams job, I'll let them know where they 
are failing in that task.  BTW part of the reason I take this hard line build 
directory separation is that when I am working alone, it keeps me from making 
too many mistakes.

>  This method works for me better than tagging first 
>  then checking out based on a tag.  
>  Utlimately, shouldn't the files in a working folder 
>  which contributed to a successful build be tagged?  

Which you can not guarantee if you did not check the tag out into a clean 
directory, i.e., there may be a file in your sandbox that is working and not 
yet in CVS.

>  I suppose either way results in the same outcome but the latter, 
>  as I described, 
>  leaves too much room for mistakes in my case it involves more work.  
>  Technically, if you tag a module in the repository first then check 
>  out based on the tag, what happens, 
>  does CVS take a quick snapshot of the repository state and then tags it 
>  and during this time no one can check out and repository 
>  transaction would be slightly slower than normal?
>  

My script below does make one assumption about my group's work process:
1) on release days the head of the trunk or branch which we are releasing 
from, *SHALL* be:
        up to date,
        compilable,
        as we intend to release.
2) which has the following set of changes to work process:
If you are not involved in the code that is getting released, you'll either 
keep it in your sandbox or you'll make your own [sub]branch to work while we 
cut the release.
If you are involved in the code that is getting released,
        you'll check in only working code in the release trunk|branch
	you will not experiment or work on future features while in the release 
trunk|branch.


> > Date: Tue, 24 Feb 2009 11:59:09 -0500
> > From: address@hidden
> > To: address@hidden
> > CC: address@hidden
> > Subject: Re: Script Help
> >
> > Rez P wrote, On 02/23/2009 07:41 PM:
> > > I'm curious as to how other build engineers keep a log, 
> > > do you run cvs stat or cvs log also?
> > > Is there a cvs command to print only the path and revsion
> > > number without resotry to scripting?
> > I assume you are trying to have a method to pull back from CVS exactly what 
> > was built for a release.
> >
> > 1) use tags, they will make your life easier and sanity last longer.
> >
> > 2) use the following method:
> > cd /tmp/
> > cvs checkout baselinename
> > cvs tag build_tag_bla baselinename
> > cvs release -d baselinename
> >
> > cd /where/you/usualy/make/release/builds
> > cvs checkout -rbuild_tag_bla baselinename
> > #and for your method of sanity, even though it is no longer needed.
> > cvs -f stat | grep "Repository revision" | gawk '{print $4 " " $3}'
> > cd baselinename
> > make\ant\buildsystem world
> >
> > sleep better. :)
> >
> > I have posted partial scripts to this list in the past for doing part of this 
> > work.
> > --
> > Todd Denniston
> > Crane Division, Naval Surface Warfare Center (NSWC Crane)
> > Harnessing the Power of Technology for the Warfighter
>
> _________________________________________________________________
> Windows Live™ Hotmail®:…more than just e-mail. 
> http://windowslive.com/explore?ocid=TXT_TAGLM_WL_t2_hm_justgotbetter_explore_022009


--
Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane)
Harnessing the Power of Technology for the Warfighter