On 8 November 2012 15:41, Ray Mansell <[email protected]> wrote: > In addition to IEBUPDTE, you should include z/VM, where XEDIT and the > CMS UPDATE command also make use of sequence numbers. > > Several decades ago, I wrote a code development library system based > around these commands, and z/VM development still uses similar > technology to develop and maintain the product.
And we also use an update identifier just before the sequence number (XEDIT can put it there). This leaves very little room for the code (and a good excuse to have sparse remarks on the code). Now that disk space is less of a concern, we're moving the sequence number further to the right for REXX code for example (like record length of 97 to leave 80 bytes for the program source). One of the virtues of CMS UPDATE is that you can actually have different code streams share the updates, as long as they don't bite each other (like when you work on a new major release, still being able to make changes to the current level). I can't think of strong enough punishment for those who came up with the idea to manually put a "change history" in the source code which normally makes it completely impossible to take an update out... Source management tools should keep track of which fixes are applied (and not rely on the programmer who might forget the manual logging). When needed, a build process can generate audit trails in the object code (like we put the list of all update files with time stamps into the object code). </soapbox> Rob
