Gabor Szabo a écrit : > On Mon, Apr 28, 2008 at 5:38 PM, Jesse Vincent <jesse@fsck.com> wrote: >>>> Site configuration information for perl v5.6.2: >>>> >>> 5.6.0 came out over 8 years ago. >>> >>> a: What's our policy on attempting to support it? >>> b: What's the diplomatic answer to the bug report as a result of that? >>> >> Would any list member who is still using Perl 5.6 in production please >> speak up? Ich. > I have two clients where they are still using 5.6.x and one that still uses > 4.x in some places.... I don't think either of those would report a bug in > their version of perl. They are ok with whatever they have and whenever > I am asked - they think I am an expert :-) - I tell them not to expect any > bug fixes as all those bugs were probably fixed in 5.8.x or 5.10.x. Yep, all my timesheet, invoicing, financial analysis, reporting and more at work is all written using 5.6.2. There is a *lot* of code. I have an open ticket to upgrade stuff to 5.8 (which I suppose I should update to 5.10 thereby leap-frogging a generation), but until some showstopper comes along (e.g. a new must-must-must-have module that doesn't work) then it's never going to happen. Then again, I know how to deal with the issues involved to upgrade, or more likely, work around any eventual 5.6 issues that may arise from here on out. In that case I wouldn't file a bug report for fear of ridicule :) > One of them showed me a 50 page long migration document. > As I can see it is mostly perl58delta with embedded comments on what > might be relevant to them. > > I thought it might be nice to have a document with some pointers on how to take > on such a migration project but then it boils down to > > - write (unit/integration/whatever) test to your code/application > - use Devel::Cover to make sure you have a high coverage of your code > - start running the same tests on the newer version of perl and start fixing the > issues that come up. > > Of course it is not that simple as most likely the old code is a huge mess > of spagetti so writing tests is very difficult and fixing lots of copy-pasted > code is very boring. Now why does that strike a chord?Thread Previous | Thread Next