largefiles: don't create chain of __contains__ calls Matt Harbison pointed out that my recent 2720f967a7b1 might cause __contains__ to continously get replaced by another version that calls itself, since the manifest instance returned by the super method is always the same instance due to @propertycache. He also suggested replacing the class instead, as is done with the context class in the surrounding code. Do so.

lazymanifest: don't depend on printf's 'hh' format to work Where we convert a 20-byte binary to a 40-byte hex string in lazymanifest_setitem(), we use sprintf("%02hhx", hash[i]). As Matt Harbison found out, 'hh' seems to be ignored on some platforms (Visual Studio?). If char is signed on such platforms, the value gets sign-extended as it gets promoted into an int when passed into the variadic sprintf(). The resulting integer value will then be printed with leading f's (14 of them on 64-bit systems), since the '2' in the format string indicates only minimum number of characters. This is both incorrect and runs the risk of writing outside of allocated memory (as Matt reported). To fix, let's cast the value to unsigned char before passing it to sprintf(). Also drop the poorly supported 'hh' formatting that now becomes unnecessary.

bundle2: test hooking using the new transaction-level hook There is no strong reason to keep a bundle2-level hook as we can use the transaction-level hook.

hook: add a generic hook right before we commit a transaction We are adding a 'txnclose' hook that will be run right before a transaction is closed. Hooks running at that time will have access to the full transaction content through both 'hookargs' content and on-disk reading. They will be able to abort the transaction.

transaction: add a validation stage The 'transaction' object can now be fed a 'validator' function. This function will be run right before the transaction is closed to validate its content. The target usage is hooks. The validation function is expected to raise an exception when it wants to abort the transaction. This only introduce the idea with a default no-op validator. Actual usage is in the next changeset.

hook: add a generic hook after transaction has been closed We are adding generic hooking for all transactions. We may have useful information about what happened during the transaction, user of the transaction should have filled the 'hookargs' dictionnary of the transaction. This hook is simple because it has no power to rollback the transaction.

hook: have a generic hook for transaction opening We are adding generic hooking for all transactions. We do not really have any useful information to include when opening the transaction but this is a useful time to allow a hook anyway. We better let people abort transaction before they happen than after multiple seconds/minutes of processing.

templater: allow piping generator-type function output to filters Template functions use "yield"s assuming that the result will be combined into a string, which means both "f -> str" and "f -> generator" should behave in the same way. Before this patch, piping generator function resulted in a cryptic error. We had to insert "|stringify" in this case. $ hg log --template '{if(author, author)|user}\n' abort: template filter 'userfilter' is not compatible with keyword '[(<function runsymbol at 0x7f5af2e8d8c0>, 'author'), (<function runsymbol at 0x7f5af2e8d8c0>, 'author')]'

record: add new tests for commit interactive (same tests as record)

record: add interactive option to the commit command

manifest: don't let find() look inside manifestdict The find() method is currently implemented by looking inside the _lm field of the manifest dict. Future manifests types (tree manifests) may not have such a field, so add a method for getting to the data instead.

largefiles: replace manifestdict.__contains__, don't extend class We're soon going to add an alternative manifest class (treemanifest). Rather than extending both those classes by largesfiles versions, let's replace the method on the manifest instances.

files: split reusable implementation into cmdutil for subrepo support

files: use ctx object to access dirstate This allows the cmdutil method to not need to be passed the repo as well as the ctx.

copies: added manifests to computenonoverlap Commit 30219bd46ed7 changed the computenonoverlap api's to not require the manifests. We actually need the manifests in the remotefilelog extension so we can find the file nodes for the various files that change. Let's add it back to the function signature with a note explaining why. This doesn't affect any behavior.

record: move dorecord from record to cmdutil Part of a serie of patches to move record from hgext to core

record: remove dependency on hg module in record Part of a series of patches to move record from hgext to core

record: remove dependency on extensions module in dorecord Part of a series of patches to move record from hgext to core

record: move filterpatch from record to patch Part of a series of patches to move record from hgext to core

patch.internalpatch: add a default value for prefix Rev 60c279ab7bd3 adding a parameter here. This breaks third-party extensions like crecord and also makes the issue fairly hard to fix on the extension's side if it wants to retain compatibility across Mercurial versions -- in old versions, the positional argument will be passed into the next unknown argument, which is 'files'. The patch also undoes a change to the record extension that is no longer necessary.

histedit: remove extra histedit constructor call In a previous commit we removed the extra histedit object instance being constructed in --continue and --abort. The new --edit-todo missed this fix though (which means the state object it produces doesn't have the locks on it). It's not breaking anything now, but let's go ahead and clean that up before we forget.

color: consolidate cut-and-paste code This fixes a mild case of cut-and-paste code regarding failing to set terminal modes. This is evident in the win32 comment that is misplaced for the terminfo mode since cset ad6ad51cc0dd. Instead, we refactor this C&P into a small local function.

record: move parsepatch from record to patch Part of a series of patches to move record from hgext to core

record: move scanpatch from record to patch Part of a series of patches to move record from hgext to core

record: move hunk class from record to patch Part of a series of patches to move record from hgext to core

record: move countChanges in the hunk class Part of a series of patches to move record from hgext to core

record: move header class from record to patch Part of a series of patches to move record from hgext to core

cmdutil.tryimportone: allow importing relative patches with --bypass

cmdutil.tryimportone: allow importing relative patches into the working dir This makes hg import --prefix dir/ (without bypass) work.

commands.import: accept a prefix option The --prefix option is meant to be relative to the root rather than the current working directory. This is for consistency with the rest of 'hg import' -- paths in patches are otherwise considered to be relative to the root. In upcoming patches we'll hook this option up to the patch functions. --exact with --prefix is currently disallowed because I can't really come up with sensible semantics for it, especially when only part of the patch is preserved.