Summary:
The original code had a very naive parser for the output of handle.exe (basically a whitespace split and an abort if a line with 4 works was found).
This diff introduces a regexp parser that will skip unknown lines.
Note: It introduces the new parser but it doesn't yet use it (left for the next diff in the stack).
This is part of a split of D56490217
Differential Revision: D56593479
fbshipit-source-id: 77c2262f18afc791efb5fdc6b4403a24fd03d33c
Summary: this diff adds support for thrift in the OSS client build. it changes the `getdeps` airstore manifest to gain access to the thrift cpp2 compiler, then uses a rule similar to other projects to generate the cpp type headers. It changes the OSS `libbundle` library to depend on the thrift types introduced in D56578073 (and, through a lot of pain, get generated in the OSS build stack) which will then be used to ser/de bundles in the OSS build later in the stack!
Reviewed By: ASchneidman
Differential Revision: D56578071
fbshipit-source-id: 1a8d906d0f7a24da21fe7f225fe68644c6bc49d7
Summary:
This diffs makes our backsyncer scuba logging log much richer data for each bookmark update log entry synced, including:
* number of blobstore gets and puts
* number of SQL queries
* number of polls for the syncing future
* time spent polling that future
Differential Revision: D56703561
fbshipit-source-id: 484bc78bd26d4ce3a21b808586c0127d3d81c124
Summary:
This refactor will alow me to log data based on aggregate runtime and other stats of the function.
NOTE: for now I'm keeping two different log tags that we were already logging
from inside function - hence the mutable tag value. I might change that later - but for now I want to make
single change per diff.
Differential Revision: D56703557
fbshipit-source-id: 4b2cc82085b319d2b9ef3010a5184c35258e4dba
Summary: I want to collect some statistics on how long does it take to backsync a single entry. For that I'd like to have a future that would represent that - this stack is doing that refactor.
Differential Revision: D56703560
fbshipit-source-id: 23713a85c5627b81a7c1f419270cb1c857818863
Summary: This config has not done actually anything since D35300521, so we can safely get rid of it
Reviewed By: zzl0
Differential Revision: D56659194
fbshipit-source-id: 9b918d77dde87c19cfbcd8e6af2343d4435360c1
Summary: The Git repos on Mononoke GRit server are deep sharded, i.e. only a subset of repos are present on any one server. This is why we need a routing layer that identifies the correct host for an incoming request for a Git repo. This crate is meant to serve that purpose. It will be used in Metagit Location Service and Metagit Follower during replication
Reviewed By: mitrandir77
Differential Revision: D56671042
fbshipit-source-id: c75c360e41a7d461fb0cdb63ba662d9133714b69
Summary: will be fixed in the following diff
Reviewed By: sggutier
Differential Revision: D56681839
fbshipit-source-id: 4e535c304c8b569f44b6e93d5de8f6619141d077
Summary:
One scenario we found during the first test merge using submodule expansion is that a user removes a submodule, but keeps a static copy of it in the same path.
This is currently failing validation, because it expects the entire expansion to be removed if the submodule metadata file is removed.
I synced with mitrandir77 and markbt about this and we got to the conclusion that this validation logic **is incorrect**.
The change made by the user is valid, so we should treat the removal of the submodule metadata file as a valid change. Everything left in the submodule path will be treated as regular file changes.
One issue I expected to see from doing this would be later changes being made to this static copy failing validation because the commit was modifying a path in the submodule dependency map (i.e. theoretically a submodule expansion) without modifying the metadata file.
So I added this scenario in these tests, because they are addressed in the fix on the next diff (D56690832).
Reviewed By: mitrandir77
Differential Revision: D56633699
fbshipit-source-id: 55c558fae26dd0ef9671b4016d43b48a3c929eba
Summary:
There are scenarios where the expansion of submodules might generate multiple file changes for the same path.
These can fall into two cases:
1. One is from the original bonsai (i.e. from small repo) and another was generated
2. Both of them were generated: one deletion and one change.
We want to make sure that we keep track of which changes were from the original bonsai and which were generated, so we can set the correct order of precedence when creating the final `SortedVectorMap`.
File changes from the original bonsai **always take precedence**. If both of the file changes were generated, the regular change should take precedence over the deletion.
## This diff
Introduces an enum that we use to make sure that original and generated file changes are kept separate and are merged together with the correct order of precedence.
Reviewed By: mitrandir77
Differential Revision: D56693920
fbshipit-source-id: c0de7ae24d67e944350f51c64af1c6bffe207795
Summary:
We're not properly handling deletion of recursive submodules.
Their submodule metadata (i.e. pointer) file has to be deleted as well and this is fixed in the next diff.
Reviewed By: mitrandir77
Differential Revision: D56528256
fbshipit-source-id: 83436404276bd83676e96bd43d7ed638739bb86d
Summary:
The scuba logging in backsyncer is lacking testing, in fact it's not even configurable rn.
This change:
* makes the scuba dataset a binary argument so we can override it (while keeping the current hardcoded value a default)
* shows the current behaviour in tests (note that some fields we're trying to show there are missing - will be added later in the stack)
Differential Revision: D56703559
fbshipit-source-id: c21f0c28fcd6c408ac031cbdf79d987a1da01c0d
Summary:
The original code considers only "fixable" the error code "used by another process", but not "access denied". However, this is the error code we get when the file we want to delete is a binary that is running.
This diff accepts that error code as fixable and therefore will trigger the kill process branch.
This is a subset of D56490217 which I've been asked to split in parts.
Reviewed By: MichaelCuevas
Differential Revision: D56589592
fbshipit-source-id: 6474ea50023131297ade949a8b166cad5cade60e
Summary: This can be simplified using `?`.
Reviewed By: YousefSalama
Differential Revision: D56700084
fbshipit-source-id: 4bac2b7851698640610de1a531b28f80615f51bc
Summary: To reduce duplication, because it will also be used in next diffs.
Reviewed By: andreacampi
Differential Revision: D56688281
fbshipit-source-id: 9b2d361413e2b80dc710d21ed7edfbf0c48d77a2
Summary:
It will be very useful to know how much time we're spending on each part of the submodule expansion and validation flow.
It will help determine if and where we need to work on improving performance, because the inital import can run for a long time, but the same code will be used by live forward and backsyncer, so it needs to be reasonably fast.
Reviewed By: andreacampi
Differential Revision: D56561379
fbshipit-source-id: 26544914299e2ef022d472a0269920fe26f5b1f4
Summary: This diff is the result of running a one-off script that modifies all usages of both `prod_entitlement_name` and `dev_entitlement_name` where they are set to the same entitlement, replacing it with the `entitlement_name` parameter. Helper functions that call `cogwheel_test` etc are ignored.
Reviewed By: jermenkoo
Differential Revision: D56633248
fbshipit-source-id: e19b49be02a014cfb27ec0fe3081330fd4baa65b
Summary:
improve format of version to Mononoke
The version reported is corresponded to either EdenFs or Sapling release.
Keep the identical prefix causes confusion.
Reviewed By: quark-zju
Differential Revision: D56580507
fbshipit-source-id: 1bac031ab8f7e60c6f18ec235ab94432b53a0922
Summary: Add a manifest for libiberty so that folly open source builds pick it up and use the demangler when showing stack traces
Reviewed By: dmm-fb
Differential Revision: D56679872
fbshipit-source-id: 622090112e19a5dd0d2b58a36dead0ea81e89026
Summary: Add a manifest for libunwind so that folly open source builds pick it up and show stack traces
Reviewed By: dmm-fb
Differential Revision: D56679859
fbshipit-source-id: 1fd3321ba1e80c4569b870974737be7bc65b77b9
Summary:
The continue and abort buttons were "icon" style buttons, unless the continue button was enabled, then it was primary.
But with the new "Configure External Merge Tool" and "Auto run merge drivers" buttons, it makes "Abort" harder to differentiate quickly.
I think making the two primary actions not use "icon" style makes them easier to see among the other buttons now.
We may revisit this again in the future, but for now I think this is an improvement.
Reviewed By: quark-zju
Differential Revision: D56658486
fbshipit-source-id: 3e23bb901af4582f901c6ad12ef65775a26038e3
Summary:
The config abstraction supports "layer"s that form a tree.
The feature is not yet utilized. This diff adds a command to show the tree
structure.
Note the `ConfigSet::layers` implementation does not include itself as a layer
so this command misses configs loaded from filesystem.
Reviewed By: muirdm
Differential Revision: D56592493
fbshipit-source-id: b9d790f97f34d10c3794bff4d58ff814df007caa
Summary: I broke this by not defining `sample` in the OSS build. Since `sample` is not used in wrapper_main() in the oss build, we also need to add `#[allow(unused_variables)]` to the function to avoid compiler warnings/errors.
Reviewed By: jdelliot
Differential Revision: D56652907
fbshipit-source-id: 821d620e7584edc75146ce735719d17fbdd49c03
Summary:
Resolves the `cargo check` issue:
error[E0599]: no method named `working_copy` found for mutable reference `&mut repo::Repo` in the current scope
--> lib/clidispatch/src/dispatch.rs:425:39
|
425 | let mut wc = repo.working_copy()?;
| ^^^^^^^^^^^^ method not found in `&mut Repo`
Reviewed By: sggutier
Differential Revision: D56592489
fbshipit-source-id: dba9077630bd56f2711a3ed30580f3901a1d4c79
Summary:
Internally, config tracks "source" at byte offset level. But that is not
friendly for end-users. Update "source" to provide a line number.
Reviewed By: sggutier
Differential Revision: D56592496
fbshipit-source-id: 2a67922abf811b8d08d5cff1285ae5bd7ed00850
Summary:
This should stop `git fetch` (triggered via the gitstore layer) from polluting the stderr and integrate with `IO`
and pager well.
Reviewed By: sggutier
Differential Revision: D56610228
fbshipit-source-id: 3cda83e6671ad6ea7fe081d12882cdb4f54d4f57
Summary:
Thrift codegen can produce bad code, like:
error[E0583]: file not found for module `mock`
--> ...\cargo-target\debug\build\fb303_core_clients-b3d87b870456fabb\out\lib.rs:9:1
|
9 | pub mod mock;
| ^^^^^^^^^^^^^
|
Do not treat it as failure.
Reviewed By: muirdm
Differential Revision: D56607940
fbshipit-source-id: 23d3e0602085f01f6fd5ed195589cd46c66124fc
Summary:
Allow websockets to connect to the current browser's port of choice (if available)
Adds a check for the current window's protocol. Defaults to ws:// but will switch to wss:// in secure contexts. This assumes the upstream server will handle converting from wss to ws.
Pull Request resolved: https://github.com/facebook/sapling/pull/886
Test Plan:
Import github project to replit. Run a server by doing:
`sl web --no-app`
open the replit webview with the token, see if it loads the smartlog.
Reviewed By: quark-zju
Differential Revision: D56526564
Pulled By: evangrayk
fbshipit-source-id: d241e2f2bfc7fb3fb30d9be9e8206970b8d57cd3
Summary:
Add a config flag to control if you want to run `sl resolve --all` before `sl continue`. This runs automated merge drivers to regenerate generated artifacts.
Sometimes `sl continue` would complain, "you have generated artifacts, run `sl resolve --all` to regenerate". We should just do this automatically for you. We provide a setting so you can opt out of the behavior. That might be useful since merge drivers can be expensive sometimes.
I support we should be doing this in sapling itself? Maybe via a flag when called from ISL?
If we do as a separate command, you'll see if it's `sl resolve --all` that's taking a long time when running in the UI (as opposed to `sl continue`). So we could add special UI documentation that shows that explains it's running merge drivers.
For OSS, I think merge drivers may be less common, so I made this false by default. We could reevaluate this later, and anyone who hasn't tweaked the setting would get the new default.
I also made an inline setting checkbox next to the continue/abort buttons to control this. But only for internal users, since it would likely just be clutter for OSS users. If we determine from analytics that everyone always leaves this active, we could just leave this buried in settings and not show it inline.
Reviewed By: quark-zju
Differential Revision: D56642944
fbshipit-source-id: a8ad580eaf98a618360c92b65d46a0101da4ba31
Summary: This feature was previously not supported, and it's a nice to have
Reviewed By: muirdm
Differential Revision: D56598268
fbshipit-source-id: 29d8247fef5841f5feae5b16ba1b81055befa8dd
Summary:
Splitting D56490217
This one makes "failed_to_kill" a list instead of a string as we want to log all processes that we couldn't kill instead of one (currently we give up on first fail, but I believe since the user will have to do some manual work we might as well do as much work as we can).
(note: change on behavior other than this will happen in the last diff of the stack)
Reviewed By: jdelliot
Differential Revision: D56588913
fbshipit-source-id: 596a0ad34e7f0e75e93c418aeba7515c80d51748
Summary: In `xdb.commit_cloud` we store CSIDs hex-encoded. These functions add a little abstraction so we can run the same query but slighly different depending on wheter we're using a mysql database or an sqlite one.
Reviewed By: markbt
Differential Revision: D56586031
fbshipit-source-id: 29237d79fc31b177b48d314003d9eb3aa289e442
Summary: The commit cloud service needs some extra operations when we execute queries so we're backwards compatible with the existing db. E.g commits are hex encoded in `xdb.commit_cloud`, so when retrieving data we need to be able to decode them to then pass them up the service.
Reviewed By: markbt
Differential Revision: D56193597
fbshipit-source-id: b45e233a09ab581d6aed7a6193e864fae9faf3f1
Summary: Most endpoint are unimplemented rigth now. Will fill them out little by little
Reviewed By: markbt
Differential Revision: D56042552
fbshipit-source-id: 2dbeb2ed7c8ce339b7c00f7b7ca8353319150540
Summary: When debugging a slow `hg st`, I saw that a lot of time was spent evicting items from the `BlobCache`. This adds the ability to turn blob caching off. This mirrors the same option from `TreeCache`
Reviewed By: MichaelCuevas
Differential Revision: D56045141
fbshipit-source-id: 8c401bf9b9b302352fa5a85365b1e10838ef86a8
Summary:
If you drag & drop a commit to rebase, we want the drag target to be a bit forgiving. Currently, it's only EXACTLY the title. So if you dragged onto the diff badge or buttons, it wouldn't work.
This often feels incorrect, for example if you have a commit with a long title, you drag from near the end of the title upwards, near remote/main, it won't preview the rebase unless you get exactly onto the bookmark name.
Instead, we can extend the non-layout part of the <DragToRebase> so that it recieves dragEnter events.
This is done the same way we support dragging from the avatar, with a positive padding and negative margin. This keeps layout the same, but increases the area available for mouse events.
Note that we need to be careful doing this to make sure this new area is not stealing mouse events that were intended for buttons or other UI elements. We can do this with z-index, which seems to work pretty well in my testing.
Reviewed By: jakebolam, sggutier
Differential Revision: D56649406
fbshipit-source-id: 7f091ba4cbd3a675597ca2da370dcd3626657d4a
Summary: This is the GC that runs in the background on Windows and in `eden doctor`, we've had a user request to manually run this, but rather than having to run `eden debug thrift`, this adds this endpoint as its own debug command for ease of use, and because the `age` parameter is silghtly confusing.
Reviewed By: jdelliot
Differential Revision: D56478945
fbshipit-source-id: d46f38d7626af5753da8db2111af4aa4d5876755
Summary:
The hint was firing when just running "sl". This was because the Rust dispatcher was setting Python's `sys.argv` to its args post-processing (which includes adding the default command "sl" on to the list of args). This made it impossible for Python to know if the user typed "sl" or "sl sl" (always appearing as the latter).
Fix by initializing the Python interpreter with the original args. We still pass the processed args to `sapling.run()`, so this should only affect the value of `sys.argv`.
Reviewed By: evangrayk
Differential Revision: D56640367
fbshipit-source-id: 43f2e20edccc54750a170319f806ad8dc79d9ca9
The internal and external repositories are out of sync. This Pull Request attempts to brings them back in sync by patching the GitHub repository. Please carefully review this patch. You must disable ShipIt for your project in order to merge this pull request. DO NOT IMPORT this pull request. Instead, merge it directly on GitHub using the MERGE BUTTON. Re-enable ShipIt after merging.
Summary:
This is done in preparation for D56220000. In all of our new kind of tests,
`modernclient` and `newclientrepo` should be used instead of the `eagerepo`
from `tinit.sh` since this function causes a number of issues when used.
Additionally, a relatively new feature, updating on clone, is tested here.
Reviewed By: muirdm
Differential Revision: D56220003
fbshipit-source-id: bb54a016b382e2eec138618b7e8c9118a3fb14cf
Summary:
On .t tests `drawdag()` allows creating a commit graph and at the same time
create bookmarks there. If we start having many tests that clone eager
repos and want to easily push all of our bookmarks there, just like we do in
D56220003, allowing to do both steps in the same bash function should be
easier.
Reviewed By: muirdm
Differential Revision: D56220002
fbshipit-source-id: 097e7624e5e958e240136358ea45d3a61a285d86
Summary: It turns out that the bash echo command prints escaped newlines as actual newlines on bash; we need this kind of behavior in D56220002, so let's fix that
Reviewed By: muirdm
Differential Revision: D56598270
fbshipit-source-id: de590fbf1262e7d35d2000eea434f124611a6608
Summary: Use <RadioGroup> instead of <VSCodeRadioGroup> for the cwd selector in the cwd dropdown
Reviewed By: muirdm
Differential Revision: D56499709
fbshipit-source-id: fbb4d1109aa67a759a3d0d4f32f9729dd04d14f4
