Skip to content

Tags: esiebomaj/pyre-check

Tags

v0.9.7

Toggle v0.9.7's commit message
Make parent file / directory tests robust to existing files

Summary:
I created files `/tmp/a` and `/tmp/b` when working on some ocaml
tests yesterday, because I wanted to diff some json output. This
broke my unit tests!

It turns out that the existing `find_parent_directory_containing_*`
logic is not possible to test robustly in the current form - it
does a recursive search of all parents of a temp directory, but
we have no control at all on the contents of directories outside
of the ones we control.

So, I tweaked code to
- allow us to optionally truncate the search at some path
- make two levels of directories in the tests
- stop search after checking the outer level

Reviewed By: grievejia

Differential Revision: D31800614

fbshipit-source-id: 6e20e8074699f9ac13ea7e8ad3a0dc3b093d8877

v0.9.6

Toggle v0.9.6's commit message
Support sanitizing all parameters in the analysis

Summary: This implements all parameters sanitizers and adds documentation for it.

Reviewed By: pradeep90

Differential Revision: D30948252

fbshipit-source-id: a503d8845a2a27cb046d8233298374eb931006f4

v0.9.4

Toggle v0.9.4's commit message
Line-level coverage information match file-level statistics (1/3)

Summary:
### Context

We already surface file-level coverage information, this is collected by  AnnotationCountCollector in statistics_collector. We are now interested in providing finer-grained coverage information that is line-level, this is collected by CoverageCollector in coverage_collector.

The existing implementation of CoverageCollector was a placeholder one with a trivial implementation. This stack of diff is intended to bring the line-level coverage information match the file-level statistics that we collect.

### This diff
As a first step towards it, collect information in CoverageCollector in a fashion similar to AnnotationCountCollector, but more granularly. So instead of counting, we are actually collecting the items that we were counting in AnnotationCountCollection. For testing, we verify that the lengths of the items we are collecting are the same as would have been with AnnotationCountCollector.

### Stack-to-be

The diffs in the stack that will follow will do the following:
1. Change the `covered_lines` and `uncovered_lines` be based on globals, attributes, parameters, etc.
2. Refactor CoverageCollector and AnnotationCountCollector, so that we have base class AnnotationCollector and share more code. We intentionally do not do this first, so as not affect statistics jobs.

Reviewed By: grievejia

Differential Revision: D30622246

fbshipit-source-id: 05479efe0eb2867959cb7d39e6f6606127abe9c4

v0.9.5

Toggle v0.9.5's commit message
Clean up another c qualifier error

Summary:
Filenames are `const char *`, not `char *` and the `statfs` interface
respects this; we should as well.

Reviewed By: grievejia

Differential Revision: D28464007

fbshipit-source-id: 10cbe9875d9f036a075a7b0ca09834170555a431

v0.9.3

Toggle v0.9.3's commit message
Update Pysa quickstart

Reviewed By: gbleaney

Differential Revision: D28416629

fbshipit-source-id: 8ebc243adc0a7e2bcd9823da34216724146c2de2

v0.9.2

Toggle v0.9.2's commit message
Reduce verbosity of overlapping buck target warning

Reviewed By: shannonzhu

Differential Revision: D28342186

fbshipit-source-id: 8c35b8c36ff60c305d01e3510b8b865ad0a6c005

v0.9.1

Toggle v0.9.1's commit message
Fix up comments on Define.Signature.t

Summary:
We had inline comments (by that I mean trailing comments) for fields in
Define.Signature.t, and in a refactor the autoformatter moved our comments for
us to the next line. This is bad because comments should always go above the
code, not below, and it was actually kind of disastrous in this case because
`parent` and `nesting_define` - the very things we wanted to the comment to
clearly distinguish - are right next to each other so the comment became a lie.

Anyway, patching up the comments to be above the record fields they refer to,
and also to be consistent across the .mli file, the signature, and the struct.

It's probably not a good idea for us to make too much use of inline comments
for this reason - autoformatting can easily drop them to a new line at which
point they go from being helpful to actively misleading.

To be honest I think this is an issue in the specification of `ocamlformat`,
moving an inline comment down is probably almost never the right thing to do.

Reviewed By: grievejia

Differential Revision: D28151440

fbshipit-source-id: 52dff080c158db0f54dbd6ddbba93307524d2dea

v0.9.0

Toggle v0.9.0's commit message
More actionable message for "server already exist" error

Reviewed By: shannonzhu

Differential Revision: D26838895

fbshipit-source-id: 978e5ed739267ed3f3388961e1cbb47f9fdf8635

v0.0.63

Toggle v0.0.63's commit message
improve error messages when loading from saved state fails

Summary: As titled.

Reviewed By: grievejia

Differential Revision: D26836065

fbshipit-source-id: 6330a1e1701035d04e9a9bc99be5a0d2187d3473

v0.0.62

Toggle v0.0.62's commit message
Resolve concatenation of tuples via unpacking.

Summary:
Problem: We used to resolve `(1, *(2, 3))` as `Tuple[int, unknown]`. This was because we correctly resolved `(2, 3)` as `Tuple[int, int]` and then dropped the ball on the `Starred` expression because there is no type to express a naked unpacked tuple.

With variadic tuples, we can handle different kinds of unpacked tuples.

Reviewed By: dkgi

Differential Revision: D26821042

fbshipit-source-id: 7aad420ca4b2775e32617077a4f1d0e965efff30