On Fri, Apr 11, 2014 at 9:19 AM, Stephan Beal <sgb...@googlemail.com> wrote:
> On Fri, Apr 11, 2014 at 6:11 PM, Matt Welland <estifo...@gmail.com> wrote: > >> 1. It should be possible to "fossil open" only a subset (or none), e.g. >> "fossil open -pop src,lib" >> 2. All operations would need to respect the populate mask (particularly >> sticky would be things like addremove) >> 3. A judgement call would have to be made on reporting commands. Should >> everything be reported (i.e. the depop'd dirs treated as if virtually >> there) or should the reports limit only to the populated files? >> > > There are internals which will fight you on this, namely the population > and traversal of the vfile table. If partial checkouts are possible, it's > possible that sub-checkouts A and B are based off of different root > versions, and vfile generally (with few exceptions) does not like there to > be more than one version's worth of information in vfile at a time (and it > likes all files in the current version to be in vfile). This would not only > be a large effort in fossil (lots of commands use vfile) but would require > a significant effort in libfossil :/ (where we're now manipulating vfile > with wild abandon!). > Can you explain what you mean by sub-checkouts A and B? I may not have been adequately clear in my description. If A and B in your paragraph refer to directories then in the model I'm proposing they would never be from different nodes in the timeline. The repo continues to be a single contour of files. I would NOT want the ability to get directory A from ab14 and directory B from 1ac3 for example. If you are working on the tip which was say 1ac3 then any action taken would *pretend* that all files in the working area were from that same node. To illustrate further, a fossil diff would NEVER show a diff for files in depopulated directories. By definition those files would be up to date and unmodified. I am not proposing a mechanism that would allow mixing and matching of content from different versions in different directories. If a project needs to do that then they should be using separate fossils and possibly using -nested. The single contour model is sacrosanct in my mind, it is a hard requirement that makes things like near trivial branching and merging possible. -- Matt -=- 90% of the nations wealth is held by 2% of the people. Bummer to be in the majority...
_______________________________________________ fossil-users mailing list fossil-users@lists.fossil-scm.org http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users