On Fri, May 6, 2011 at 12:24 AM, Keshav Kini <keshav.k...@gmail.com> wrote: > Hi Simon, > > There is no 4.7 yet, so I'm not sure how the patchbot can do what you want. > Robert Bradshaw has set up the patchbot to work from the last stable > release, which is 4.6.2. You may need to rebase your entire ticket (with its > instructions for dependencies etc.) on 4.6.2, or include as dependencies any > relevant patches that were merged in the alphas or release candidates.
Actually, as of yesterday, it's building against 4.7rc1. http://sage.math.washington.edu:21100/?base=4.7.rc1 > In any case, it looks like the patchbot is applying a patch for #11115 > because #11298 depends on #9976, which the bot somehow thinks depends on > #11115. This may be due to the large number of lines in the comments > containing "#11115" and "dependency" or "depends"; the patchbot tries to > guess what people are talking about, and sometimes guesses wrong, it would > seem. > > Now that Maarten Derickx has implemented a "dependencies" field on trac > tickets, I think the patchbot should strictly use that field if it exists, > and not guess based on comment parsing - it would likely solve problems > like this. Hopefully we will eventually have some way to succinctly and > accurately express patch application orders directly on trac as well. +1, I'll be changing that. Note that it also needs to understand the "closed/merged in" fields as well to avoid double application. > I believe Robert Bradshaw's original intent was to make the patchbot > unobtrusive and work on its natural language processing abilities to improve > accuracy, but I think that as we become more reliant on it it becomes more > necessary for us to be able to control it in a more predictable way. Yes. The optimal workflow I see is 1) Work on ticket 2) If there's anything complicated going on, check the patchbot to make sure it got things "right." 3) If necessary, and hopefully not often, explicitly add correction. This will place the burden on complicated tickets rather than burdening simple tickets with hoops to jump through. One of the big points of the patchbot was to reduce the manditory O(1) overhead of creating and reviewing a ticket. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
