Community
Participate
Working Groups
there are a number of places in our scripts where we specify to use BUILD_CVS_WRITE_USER and BUILD_CVS_WRITE_PROTOCOL when the 'read' variables should be used instead. This was probably partially oversite, and partially just a result of things changing over time. For example, we should be about to use only the "read" variables in checkout.xml. The write variables are needed only in maptasks.xml, and then only if donottagmapfiles is false (i.e. we do want to tag map files). To some extent, this does not matter too much, the same protocols and ID will often be desired to use for both read and write (when not using pserver) but at least cause confusion and is not "self documenting". Ideally, the "write" versions would not have to be specified at all, unless one was wanting to tag map files.
Status: I'm made the changes to main streams now. Still not patches. "cvs write user" should no longer be required unless it is desired/attmpted to tag the map files. There is a few test runs building, so will check those over weekend, and make sure they worked as expected (and not tag the map files).
I've fixed so that BUILD_CVS_WRITE_USER is no longer used, except if/when builder wants to automatically tag map files ... but, then, only if BUILD_CVS_WRITE_USER has some value other than "anonymous". So, better documented, and at same time, no longer need a specific committer id hard coded into wtpBuild scripts ... at the sacrifice of no longer tagging each build.
I meant to add ... I put these changes into main HEAD build, which is used for juno build, indigo maintenance, and helios maintenance, and 3.2.4 patch builds (even though, doubt we'll do any more 3.2.4 patch builds). I have also incorporated the changes into the branched builder of used for the old 1.5.5 patches and 3.0.5 patch builds. That's stream v201003030847c (currently tagged as v201003030847c-2). So (the reason for this long note) these two patch builds are essentially untested. If/when a patch build is ever needed .... may want to "run early" to test out that things still build as expected.
