Summary:
With flambda (`-O3`), compilation time is ~5x slower, but the backend is ~25% faster!
To mitigate the atrocious compilation times, introduce a new `opt` build mode in the jbuilder files.
- build in "opt" mode by default from the toplevel (so that install scripts and external users get the fastest infer by default), in "default" mode by default from infer/src (since the latter is only called directly by infer devs, for faster builds)
- `make byte` is as fast as before in any mode
- `make test` will build "opt" by default, which is very slow. Solution for testing (or building the models) locally: `make BUILD_MODE=default test`.
- You can even change the default locally with `export BUILD_MODE=default`.
The benchmarks are to be taken with a sizable pinch of salt because I ran them only once and other stuff could be running in the background. That said, the perf win is consistent across all projects, with 15-20% win in wallclock time and around 25% win in total CPU time, ~9% win in sys time, and ~25% fewer minor allocations, and ~5-10% fewer overall allocations. This is only for the backend; the capture is by and large unaffected (either the same or a tad faster within noise range).
Here are the results running on OpenSSL 1.0.2d on osx (12 cores, 32G RAM)
=== base
infer binary: 26193088 bytes
compile time: 40s
capture:
```lang=text
real 1m7.513s
user 3m11.437s
sys 0m55.236s
```
analysis:
```lang=text
real 5m41.580s
user 61m37.855s
sys 1m12.870s
```
Memory profile:
```lang=json
{
...
"minor_gb": 0.1534719169139862,
"promoted_gb": 0.0038930922746658325,
"major_gb": 0.4546157643198967,
"allocated_gb": 0.6041945889592171,
"minor_collections": 78,
"major_collections": 23,
"compactions": 7,
"top_heap_gb": 0.07388687133789062,
"stack_kb": 0.3984375,
"minor_heap_kb": 8192.0,
...
}
```
=== flambda with stock options (no `-Oclassic`, just the same flags as base)
Exactly the same as base.
=== flambda `-O3`
infer binary: 56870376 bytes (2.17x bigger)
compile time: 191s (4.78x slower)
capture is the same as base:
```lang=text
real 1m9.203s
user 3m12.242s
sys 0m58.905s
```
analysis is ~20% wallclock time faster, ~25% CPU time faster:
```lang=text
real 4m32.656s
user 46m43.987s
sys 1m2.424s
```
memory usage is a bit lower too:
```lang=json
{
...
"minor_gb": 0.11583046615123749, // 75% of previous
"promoted_gb": 0.00363825261592865, // 93% of previous
"major_gb": 0.45415670424699783, // about same
"allocated_gb": 0.5663489177823067, // 94% of previous
"minor_collections": 73,
"major_collections": 22,
"compactions": 7,
"top_heap_gb": 0.07165145874023438,
"stack_kb": 0.3359375,
"minor_heap_kb": 8192.0,
...
}
```
=== flambda `-O2`
Not nearly as exciting as `-O3`, but the compilation cost is still quite high:
infer: 37826856 bytes
compilation of infer: 100s
Capture and analysis timings are mostly the same as base.
Reviewed By: jberdine
Differential Revision: D4867979
fbshipit-source-id: 99230b7
Summary: This will run infer with `--no-keep-going` by default in our tests.
Reviewed By: sblackshear
Differential Revision: D5814237
fbshipit-source-id: c1e1a4e
Summary:
Keep track of generated files better and make sure we nuke them on `make clean`.
Also get rid of a few unused variables in infer/src/Makefile.
Reviewed By: sblackshear
Differential Revision: D5754553
fbshipit-source-id: b5fca41
Summary:
This gets rid of the horrible hack of asking jbuilder to build
_build/{default,test}/.ppx/ppx_compare/ppx.exe prior to starting several
jbuilds in parallel. We now refrain from starting several jbuilds in parallel.
Also, I finally understood that order-only deps do nothing for phony targets,
and not much else for non-phony ones. I hate make even more.
Reviewed By: jberdine
Differential Revision: D5678699
fbshipit-source-id: 52179e8
Summary:
This simplifies the jbuild files: no need to list these files explicitly
anymore, nor to exclude them explicitly from the main `InferModules` library
(due to their different compilation flags).
Isolate common parts into jbuild.common do `cat`-based code inclusion into
jbuild files to factorize code.
Reviewed By: jberdine
Differential Revision: D5678328
fbshipit-source-id: 6d7d925
Summary:
The issue was as follows:
- ppx.exe depended on all the ocaml source files
- ppx.exe is an implicit prerequisit of all the (non-automatically-generated) build objects in jbuilder
- as a result, when modifying a source file, we would recompile ppx.exe, thus all the source files
ppx.exe only needs to depend on all the generated source files, so that
jbuilder can find them all.
Reviewed By: jberdine
Differential Revision: D5621007
fbshipit-source-id: 92c5b9c
Summary:
Use jbuilder to build infer instead of ocamlbuild. This is mainly to get faster builds:
```
times in 10ms, ±differences measured in speedups, 4 cores
| | ocb total | jb | ±total | ocb user | jb | ±user | ocb cpu | jb | ±cpu | ocb sys | jb | ±sys |
|-----------------------------------+-----------+------+--------+----------+------+-------+---------+-----+------+---------+------+------|
| byte from scratch | 6428 | 2456 | 2.62 | 7743 | 6662 | 1.16 | 138 | 331 | 2.40 | 1184 | 1477 | 0.80 |
| native from scratch | 9841 | 4289 | 2.29 | 9530 | 8834 | 1.08 | 110 | 245 | 2.23 | 1373 | 1712 | 0.80 |
| byte after native | 29578 | 1602 | 18.46 | 4514 | 4640 | 0.97 | 170 | 325 | 1.91 | 543 | 576 | 0.94 |
| change infer.ml byte | 344 | 282 | 1.22 | 292 | 215 | 1.36 | 96 | 99 | 1.03 | 040 | 066 | 0.61 |
| change infer.ml native | 837 | 223 | 3.75 | 789 | 174 | 4.53 | 98 | 99 | 1.01 | 036 | 47 | 0.77 |
| change Config.ml byte | 451 | 339 | 1.33 | 382 | 336 | 1.14 | 97 | 122 | 1.26 | 056 | 80 | 0.70 |
| change Config.ml native | 4024 | 1760 | 2.29 | 4585 | 4225 | 1.09 | 127 | 276 | 2.17 | 559 | 644 | 0.87 |
| change cFrontend_config.ml byte | 348 | 643 | 0.54 | 297 | 330 | 0.90 | 96 | 67 | 0.70 | 038 | 102 | 0.37 |
| change cFrontend_config.ml native | 1480 | 584 | 2.53 | 1435 | 906 | 1.58 | 106 | 185 | 1.75 | 136 | 178 | 0.76 |
#+TBLFM: $4=$2/$3;f2::$7=$5/$6;f2::$10=$9/$8;f2::$13=$11/$12;f2
50 cores
| | ocb total | jb | ±total | ocb user | jb | ±user | ocb cpu | jb | ±cpu | ocb sys | jb | ±sys |
|---------------------+-----------+------+--------+----------+------+-------+---------+----+------+---------+------+------|
| byte from scratch | 9114 | 2061 | 4.42 | 9334 | 5133 | 1.82 | | | 0/0 | 2566 | 1726 | 1.49 |
| native from scratch | 13481 | 3967 | 3.40 | 12291 | 7608 | 1.62 | | | 0/0 | 3003 | 2100 | 1.43 |
| byte after native | 3467 | 1476 | 2.35 | 5067 | 3912 | 1.30 | | | 0/0 | 971 | 801 | 1.21 |
#+TBLFM: $4=$2/$3;f2::$7=$5/$6;f2::$10=$9/$8;f2::$13=$11/$12;f2
```
Menu:
1. Write a jbuild file, autogenerated from jbuild.in because we need to fill in
some information at build-time (really, at configure time, but TODO), such as
whether or not clang is enabled.
2. Nuke lots of stuff from infer/src/Makefile that is now in the jbuild file
3. The jbuild file lives in infer/src/ so it can see all the sources. If we put it somewhere else, eg, infer/, then `jbuilder` scans too many files (all irrelevant) and takes 2.5s to start instead of .8s. Adding irrelevant directories to jbuild-ignore does not help.
4. jbuilder does not support subdirectories, so resort to listing all the
source files in the generated jbuild (only source directories need to be
manually listed in jbuild.in though). Still, the generated .merlin is wrong
and makes merlin find source files in _build, so manually tune it to get
good merlin support. We also lose some of merlin for unit tests as it
cannot see their build artefacts anymore.
5. checkCopyright gets its own jbuild because it's standalone. Also, remove
some deprecation warnings in checkCopyright due to the new version of Core from
a while ago.
6. Drop less-used Makefile features (they had regressed anyway) such as
building individual modules. Also, building mod_dep.pdf now takes all the
source files available so they better build (before, it would only take the
source files from the config, eg with or without clang) (that's pretty minor).
7. The toplevel is now built as a custom toplevel because that was easier. It
should soon be even easier: https://github.com/janestreet/jbuilder/issues/210
8. Move BUILTINS.mli to BUILTINS.ml because jbuilder is not happy about
interface files without implementations.
In particular, I did not try to migrate too much of the Makefile logic to jbuilder,
more can be done in the future.
Reviewed By: jberdine
Differential Revision: D5573661
fbshipit-source-id: 4ca6d8f
Summary:
Phony dependencies are not good for incrementality. Retain the nice
factorization of code introduced by D5415318 but depend directly on the
relevant files instead of the phony target.
Reviewed By: mbouaziz
Differential Revision: D5461188
fbshipit-source-id: 069519b
Summary:
Fixes 2 issues with the build:
- The OCaml dependencies (Version.ml, atdgen-generated stuff, ...) would be generated twice when running `make -j test` from a clean tree, because `make` has no inter-Makefile visibility over common dependencies.
- We would run `atdgen` twice: once for the .ml, once for the .mli, even though one invocation is enough for both.
Add `make ocaml_clean` to nuke only the OCaml code (in particular because
rebuilding the C++ code in the plugin is slow when all I want to get rid of the
OCaml stuff).
Reviewed By: mbouaziz
Differential Revision: D5415318
fbshipit-source-id: 16f42c6
Summary:
Conversion and reformat of infer source using ocamlformat
auto-formatting tool.
Current status:
- Because Reason does not handle docstrings, the output of the
conversion is not 'Warning 50'-clean, meaning that there are
docstrings with ambiguous placement. I'll need to manually fix
them just before landing.
Reviewed By: jvillard
Differential Revision: D5225546
fbshipit-source-id: 3bd2786
Summary:
This will be needed to re-use the functions now in Driver.ml in other contexts
without always adding to infer.ml. For instance, this is used in a later diff
to do a diff-analysis orchestrator that needs to run the capture and analysis
several times.
Reviewed By: jberdine
Differential Revision: D5319862
fbshipit-source-id: caf9551
Summary:
Once the fixed/preexisting/introduced sets have been computed, they endure
further filtering which may decide that more of them are equal. These bugs just
get dropped on the floor. Put these into preexisting as well instead, at least
in the case of the "skip_duplicated_types_on_filenames" filter.
Reviewed By: martinoluca
Differential Revision: D5274248
fbshipit-source-id: 99b3f3d
Summary:
It's easier to distinguish what's checked into infer and what comes from the
plugin this way.
Reviewed By: akotulski
Differential Revision: D5191581
fbshipit-source-id: 579311e
Summary:
This should help prevent new modules being created that do not `open! IStd`.
Documented this in CONTRIBUTING.md.
Reviewed By: jberdine
Differential Revision: D5183552
fbshipit-source-id: 0f0a8ce
Summary:
Spacetime profiling showed that this was allocating a lot more than it needs
to. Switching to a `Hashtbl` makes the memory overhead go away, and halves the
memory consumption of the whole frontend on some pathological files.
This also brings the file "clang_ast_main.ml" into the infer repo, renamed to
"ClangPointers.ml" (and given an mli). It's useful to have something like
"clang_ast_main.ml" checked into the facebook-clang-plugins repo to illustrate
how to use the OCaml AST visitor generated by atdgen, but it can be simplified
to be more pedagogical instead of being the visitor used in infer.
Reviewed By: jberdine
Differential Revision: D4884383
fbshipit-source-id: 88f324a
Summary:
Recently we changed the binaries we build and use but it wasn't obvious that
some of the old binaries disappeared. For instance, nothing short of `git clean
-xfd` would get rid of "infer/bin/InferPrint", which can lead to frustrating
attempts to make InferPrint not segfault.
Mitigate this in two ways:
- be more strict in the binaries we ignore in .gitignore, so InferPrint would should up in "git status"
- be less strict in the binaries we clean out with `make clean` so that
InferPrint et al. gets deleted by `make clean`
Reviewed By: jberdine
Differential Revision: D5120573
fbshipit-source-id: 44e7954
Summary:
Introduce `infer-<command>` for each command, except for internal commands
(only `infer-clang` for now) which are not exported. Install these executables
(which are just symlinks to `infer`) on `make install`. The main executable
looks at the name it was invoked with to figure out if it should behave as a
particular command.
Get rid of `InferClang`, `InferAnalyze`, and `InferPrint`. As a bonus, we now
only need to build one executable: `infer`, which should be a few seconds
faster (less link time).
`InferAnalyze` is now `infer-analyze` and `InferPrint` is `infer-print`. To run
`InferClang`, use a symlink named `clang`, `clang++`, etc. to `infer`. There
are such symlinks available in "infer/lib/wrappers/" already.
I also noticed that the scripts in xcodebuild_wrappers/ don't seem useful
anymore, so use wrappers/ instead, as for `make`.
Reviewed By: mbouaziz
Differential Revision: D5036495
fbshipit-source-id: 4a90030
Summary:
Glorious, glorious man pages.
This changes a bunch of things that were hard to break up from the diff, so the
resulting diff is big.
Use `Cmdliner.Manpage` to format man pages (and also format a bit ourselves
since it is so stubborn).
As a bonus, introduce the following subcommands:
```
infer run ... # same as default mode with -- before
infer capture ... # -a capture
infer compile ... # -a compile
infer analyze ... # InferAnalyze
infer report ... # InferPrint
infer diff ... # this one is not new
infer clang ... # InferClang, not that you should use it
```
The man pages can still be improved a lot. Notable missing sections:
`ENVIRONMENT`, stuff about .inferconfig, some example usage, `DESCRIPTION`, ...
Reviewed By: jberdine
Differential Revision: D4921083
fbshipit-source-id: 9602230
Summary: This will be needed to generate man pages with an accurate date.
Reviewed By: jberdine, mbouaziz
Differential Revision: D4937498
fbshipit-source-id: b5ebd31
Summary:
Went through the trouble of killing these warnings, make sure they do not creep
up again in the future.
Reviewed By: mbouaziz
Differential Revision: D4937493
fbshipit-source-id: 0e2eda4
Summary:
We won't be needing to know the terminal size when we move the help to
manpages. Hopefully if we need C stubs again in the future we can refer back to
this diff to revive them.
Reviewed By: jberdine
Differential Revision: D4937491
fbshipit-source-id: 99139fa
Summary:
We lost InferClang++ a long time ago... It was installed properly on `make
install` but not during normal compilation!
Reviewed By: akotulski
Differential Revision: D4905947
fbshipit-source-id: 56375c2
Summary:
This debugger stops the execution of CTL during the evaluation of formulas, then shows useful info about the current state:
- Which formula is being evaluated
- In which part of the AST is such formula being evaluated
- Line number of the source file
- Whether a CTL-Transition will take place or not
- Node-ID that can be related to the dotty file containing the evaluation graph (the one in infer-out/lint_dotty/)
The AST is shown with coloured nodes, where each colour has the following meaning:
- [Default terminal colour], Bold: The node is being evaluated
- Green, Bold: The formula returned true on the current node
- Red, Bold: The formula returned false on the current node
Reviewed By: ddino
Differential Revision: D4834451
fbshipit-source-id: c9aa970
Summary:
We shouldn't install byte-executables unless the user types `make byte`.
Othewise everything gets very slow and the cause is not obvious.
Reviewed By: jeremydubreil
Differential Revision: D4779135
fbshipit-source-id: 757bfa2
Summary:
It can be useful when debugging infer or the Makefiles themselves to see what
`make` is doing. Instead of editing Makefiles to remove `@` now you can `make
VERBOSE=1`.
This is just `git ls-files | grep -e Makefile -e '.*\.make' | xargs sed -e 's/^\t@/\t$(QUIET)/' -i`, and adding the definition of `QUIET` to Makefile.config.
Reviewed By: jeremydubreil
Differential Revision: D4779115
fbshipit-source-id: e6e4642
Summary:
This removes the dependency of libffi, and gives better guarantees at
compile-time. A bit overkill to retrieve the width of the terminal but what do
you know...
Reviewed By: jberdine
Differential Revision: D4700003
fbshipit-source-id: 036989b
Summary: This is dead code, and depends on libffi which we could get rid of otherwise.
Reviewed By: jberdine
Differential Revision: D4705645
fbshipit-source-id: efd28ef
Summary:
This makes sure that one can run `./build-infer.sh` then `make`. Otherwise it's
not always clear what one should do to recompile infer, eg when `make` will
work and when `./build-infer.sh` should be used instead, in particular when the
user doesn't have opam configured for her terminal.
Reviewed By: jberdine
Differential Revision: D4698159
fbshipit-source-id: 5df8059
Summary:
This is useful to make sure `./build-infer.sh clang` and `./build-infer.sh
java` work correctly.
I had to move one unit test inside a new unit/clang/ directory used only if clang is enabled, and create a stub for it in unit/clang_stubs/ for when clang is disabled.
Renamed `OCAML_SOURCES` to the more consistent and descriptive `OCAML_CONFIG_SOURCES`.
Reviewed By: jeremydubreil
Differential Revision: D4571997
fbshipit-source-id: 9502114
Summary:
It's annoying that changing the Makefile or running `./configure` with
different options does not always retrigger compilation of some targets, eg
infer.
Reviewed By: akotulski
Differential Revision: D4571988
fbshipit-source-id: dda080f
Summary:
One gets very obscure errors when trying to run infer for clang when it was
compiled for Java, or vice-versa. This diff makes sure we crash early with the
appropriate error message. For instance:
```
$ ./build-infer java
$ infer -- clang -c hello.c
Uncaught exception:
(Failure
"Unsupported build mode: make/cc\
\nInfer was built with clang analyzers disabled.\
\nPlease rebuild infer with clang enabled.\
\n")
Raised at file "pervasives.ml", line 30, characters 22-33
Called from file "backend/infer.ml", line 398, characters 6-48
Called from file "backend/infer.ml", line 449, characters 20-38
$ infer --clang-compilation-db-files foo.json
Uncaught exception:
(Failure
"Unsupported build mode: clang compilation database\
\nInfer was built with clang analyzers disabled.\
\nPlease rebuild infer with clang enabled.\
\n")
Raised at file "pervasives.ml", line 30, characters 22-33
Called from file "backend/infer.ml", line 392, characters 8-65
Called from file "backend/infer.ml", line 449, characters 20-38
```
Reviewed By: sblackshear
Differential Revision: D4566641
fbshipit-source-id: d9a118f
Summary:
- inferbo introduced a dependency to extlib. When building Java analyzers, this
is implicitly pulled in by javalib, but it's missing when building only the
clang analyzers. Add `extlib` to the packages we build against.
- infer.ml and Javac.ml depend on Javalib, but it's easy to push down the code
that needs it to `jMain.ml` so that we can build without javalib for the
clang-only case.
- jMain.mli had 2 copies: one in java/ and one in java_stubs/. Make one a symlink to the other.
Reviewed By: jeremydubreil
Differential Revision: D4566581
fbshipit-source-id: 214a4eb
Summary:
Native compilation seems to take a couple of seconds to build the unit test
binary, but we don't need it to be native (and neither do users building infer
to install it on their machines). In fact, the test mode rebuilds it in
bytecode mode. For the sake of being able to run the unit tests when developing
on infer (and thus possibly while having some fatal warnings showing up in the
code, preventing us from building the test target), also build InferUnit in
bytecode mode.
Reviewed By: sblackshear, martinoluca
Differential Revision: D4537455
fbshipit-source-id: 374c84c
Summary:
Previously, we would first compute which build command is at hand, based on the
first argument after "infer --", then do everything depending on that piece of
information. However, the build command alone is not enough to know in which
"build mode" we are operating. For instance, there are several build modes
corresponding to "buck" build commands.
This led to duplication of the logic (to retrieve which build mode we are in in
the various phases of an infer run), and some invariants that had to be
re-asserted at various points in the code, eg that the arguments are not empty.
This diff adds a `build_mode` type (renaming the previous `build_mode` to
`build_system`) that identifies the various integrations we support. We compute
the build mode at the start of infer, then pass the build mode around.
Also, move `run_javac` to a new `integration/Javac.ml` file given that it's a
bit large.
Reviewed By: jberdine
Differential Revision: D4415074
fbshipit-source-id: db854a0
Summary:
1. One call to `Core.Std.String.slice` was wrong and caused the program to crash, and
2. The crash was silently ignored because the error code of uncaught OCaml
exceptions was the same as `CheckCopyright.copyright_malformed_exit_code` (=2)
Address both issues. Also build CheckCopyright with debug options.
Reviewed By: jberdine
Differential Revision: D4410306
fbshipit-source-id: d73b086
Summary:
`make byte` will populate infer/bin/ with bytecode version of each executable,
plus infer/bin/infer.byte (used to remember which of the native or byte
executables have been built most recently). `make infer` now also creates
infer/bin/infer.native, so that we're sure to replace the executables with
native/byte versions as appropriate.
This is to make debugging a tad easier:
make byte
ledit ocamldebug $(which infer) <infer args>
Whereas previously one had to:
make -C infer/src byte
ledit ocamldebug infer/_build/infer/backend/infer.byte <infer args>
Reviewed By: jberdine
Differential Revision: D4409476
fbshipit-source-id: ab5f57d
Jules Villard
Sam Blackshear
Jeremy Dubreil
Josh Berdine
Martino Luca
Andrzej Kotulski
Dulma Churchill
Cristiano Calcagno
Kihong Heo