On Wed, Dec 23, 2009 at 10:21, Bob Archer <bob.arc...@amsi.com> wrote: >> 2009/12/23 Bob Archer <bob.arc...@amsi.com> >> > On Dec 23, 2009, at 00:45, Julian Mitchell wrote: >> > >> > > The project that I am working on utilises a code generation >> tool. >> > The header of every source file includes a comment with a >> date\time >> > stamp of when it was generated. The problem is that every time >> the >> > code is generated the svn change check algorithm marks all files >> as >> > having been changed even though only a handful have actually had >> > actual code changes. >> > > >> > > Is there a way to tailor the change check algorithm with, say, >> a >> > regex, to ignore certain contents of a text file e.g. comment >> > lines? >> > >> > To my knowledge, there is not. You could consider writing a >> client- >> > side script that committers should run before checking in source, >> > to normalize such comment lines e.g. to remove the date/time. You >> > could also write a companion server-side hook script to reject >> any >> > commit where the only difference is such a comment line. >> I think the current wisdom is, don't source control files that can >> be generated. So, for the same reason you [usually] don't store >> binaries that you build from your source don't store code files >> that are generated. Make the generation part of the build so that >> any dev running the build script gets the files generated for them. >> >> BOb >> Thanks for both your input. >> Ryan - is there a convenient place to hook in to the client side? I >> would like to catch this prior it to being displayed as a >> modification. >> Bob - you are correct and I agree with you however the generation >> process takes a while (30s - 1minute) and the controlled package >> files are ghastly to diff from a code readability perspective. >> > > Hmm... I see. Can you possibly put the results of the genned code onto a > shared location so the devs can just pull the latest down rather than needing > to take... wait 30 SECONDS?
30 seconds is nothing, really. My primary app takes 25 minutes to deploy into Tomcat whenever I'm doing a code/test cycle on my laptop. If an extra step added 30 seconds, it'd make no difference to me.