sidedata-exchange: rewrite sidedata on-the-fly whenever possible When a A exchanges with B, the difference of their supported sidedata categories is made, and the responsibility is always with the client to generated it: - If A pushes to B and B requires category `foo` that A does not have, A will need to generate it when sending it to B. - If A pulls from B and A needs category `foo`, it will generate `foo` before the end of the transaction. - Any category that is not required is removed. If peers are not compatible, abort. It is forbidden to rewrite sidedata for a rev that already has sidedata, since that would introduce unreachable (garbage) data in the data file, something we're not prepared for yet. Differential Revision: https://phab.mercurial-scm.org/D10032

revlog-index: add `replace_sidedata_info` method During a `pull` operation where the server does not provide sidedata, the client that requires it should generate them on-the-fly. In the generic case, we need to wait for the changelog + manifests + filelogs to be added, since we don't know what the sidedata computers might need: this means rewriting the sidedata of index entries from within the pull transaction (and no further back) right after we've added them. Both Python and C implementations only allow for rewriting the sidedata offset and length for revs within the transaction where they were created. Differential Revision: https://phab.mercurial-scm.org/D10031

revlogv2: temporarily forbid inline revlogs See inline comments. I plan to fix the underlying issue before revlogv2 is stabilized. Differential Revision: https://phab.mercurial-scm.org/D10030

changegroupv4: add sidedata helpers These helpers carry the information and computers needed to rewrite sidedata when generating/applying patches. We will be making use of them soon. Differential Revision: https://phab.mercurial-scm.org/D10029

revlog: add attribute on revlogs that specifies its kind The sidedata logic needs to check whether the revlog it's working on is a changelog, a manifest or a filelog. Furthermore, future versions of the revlog format will most likely see a split between the three types (i.e. they will store different information), so having this will be useful for other future endeavors as well. Differential Revision: https://phab.mercurial-scm.org/D10151

sidedata-exchange: add `wanted_sidedata` and `sidedata_computers` to repos Each repo will advertise the sidedata categories it requires (categories being unique and canonical), and have a set of "computers", functions to generate sidedata from `(repo, revlog, rev, previous_sidedata)`, for a given category. The set of computers can be a superset of the set of the wanted categories, but not smaller: repos are expected to be coherent in their handling of sidedata. Differential Revision: https://phab.mercurial-scm.org/D10028

delta: add sidedata field to revision delta When emitting revision delta, we need to also emit the sidedata information just added in the revlogv2 format if appropriate. Differential Revision: https://phab.mercurial-scm.org/D10027

changegroup: add v4 changegroup for revlog v2 exchange This change only adds the required infrastructure for the new changegroup format and does not do any actual exchange. This will be done in the next patches. Differential Revision: https://phab.mercurial-scm.org/D10026

revlogv2: don't assume that the sidedata of the last rev is right after data We are going to be rewriting sidedata soon, it's going to be appended to the revlog data file, meaning that the data and the sidedata might not be contiguous. Differential Revision: https://phab.mercurial-scm.org/D10025

sidedata: move to new sidedata storage in revlogv2 The current (experimental) sidedata system uses flagprocessors to signify the presence and store/retrieve sidedata from the raw revlog data. This proved to be quite fragile from an exchange perspective and a lot more complex than simply having a dedicated space in the new revlog format. This change does not handle exchange (ironically), so the test for amend - that uses a bundle - is broken. This functionality is split into the next patches. Differential Revision: https://phab.mercurial-scm.org/D9993

cext: add support for revlogv2 This enables the C code to retrieve/create entries in the revlog v2 format. This is mainly a matter of taking into account the additional slots for sidedata, etc. Differential Revision: https://phab.mercurial-scm.org/D9846

bitmanipulation: add utils to read/write bigendian 64bit integers Differential Revision: https://phab.mercurial-scm.org/D9845