commit: make commit acquire store lock before processing for consistency If acquisition of wlock waits for another "hg commit" process to release it, dirstate will refer newly committed revision after acquisition of wlock. At that time, '00changelog.i' on the filesystem contains this new revision, but in-memory 'repo.changelog' doesn't, if it is cached without store lock (slock) before updating by another "hg commit". This makes validating parents at re-loading 'repo.dirstate' from '.hg/dirstate' replace such new revision with 'nullid'. Then, 'localrepository.commit()' creates "orphan" revision (see issue4368 for detail). a01d3d32b53a makes 'commands.commit()' acquire both wlock and slock before processing to avoid this issue at "hg commit". But similar issue can occur even after a01d3d32b53a, if 3rd party extension does: - refer 'repo.changelog' outside wlock scope, and - invoke 'repo.commit()' directly (instead of 'commands.commit()') This patch makes 'commit()' acquire slock before processing, to refer recent changelog at validating parents of 'repo.dirstate'.

censor: make censor acquire locks before processing Before this patch, "hg censor" executes below: - without acquisition of wlock, examine whether the working directory refers the revision of the file to be censored or not - without acquisition of store lock (slock), replace existing filelog of file to be censored with censored one, Replacement consists of steps below, and it is assumed that the destination filelog at (1) isn't changed before renaming at (3). 1. read existing filelog in 2. write filelog entries (both censored and not) into temporary file 3. rename from temporary file to existing filelog to be censored It may cause unintentional result, if another command runs parallelly (see also issue4368). This patch makes "hg censor" acquire wlock and slock before processing.

transplant: widen wlock scope of transplant for consitency while processing Before this patch, "hg transplant" executes below before acquisition of wlock. - cmdutil.checkunfinished() - repo.status() for dirty check - repo.dirstate.parents() It may cause unintentional result, if another command runs parallelly (see also issue4368). This patch makes "hg transplant" acquire wlock before processing instead of acquiring wlock in each of 'transplanter.apply()' and 'transplanter.recover()'.