improve case with both anonymous lifetime parameters #43269 #43298

gaurikholkar-zz · 2017-07-17T20:34:22Z

This is a fix to #43269.

Sample output message-


error[E0623]: lifetime mismatch
  --> $DIR/ex3-both-anon-regions.rs:12:12
   |
11 | fn foo(x: &mut Vec<&u8>, y: &u8) {
   |                    ---      --- these references must have the same lifetime
12 |     x.push(y);
   |            ^ data from `y` flows into `x` here

error: aborting due to 2 previous errors

r? @nikomatsakis

estebank

Great work! A couple of nitpicks inline.

estebank · 2017-07-17T20:55:06Z

src/librustc/infer/error_reporting/named_anon_conflict.rs.bk

@@ -0,0 +1,201 @@
+// Copyright 2012-2013 The Rust Project Developers. See the COPYRIGHT


Pretty sure this file shouldn't be in the PR.

estebank · 2017-07-17T20:55:14Z

src/test/ui/lifetime-errors/ex3-both-anon-regions.stderr

+11 | fn foo(x: &mut Vec<&u8>, y: &u8) {
+   |                    ---      --- these references must have the same lifetime
+12 |     x.push(y);
+   |            ^ data from `y` flows into `x` here


estebank · 2017-07-17T20:57:01Z

src/test/ui/lifetime-errors/ex3-both-anon-regions.stderr

+  --> $DIR/ex3-both-anon-regions.rs:12:12
+   |
+11 | fn foo(x: &mut Vec<&u8>, y: &u8) {
+   |                    ---      --- these references must have the same lifetime


Not sure if having a label on only one of the arguments is good enough (but no strong opinion in any direction), given that these are also displayed on other tools, like VSCode (through RLS). I'm wondering if @nikomatsakis has an opinion about this.

Are you saying -- if some other tool were to hide one of these labels, would that be ok? It seems like it would be too bad, but I'm not sure what to do about it. I guess we could offer multiple labels?

My worry is for example, a client that displays information for each span individually without the entire context of the error on hover. I feel that this is another case of conflicting requirements for the RLS and for rustc visual output.

estebank · 2017-07-17T20:58:14Z

src/librustc/diagnostics.rs

@@ -2026,4 +2026,5 @@ register_diagnostics! {
    E0495, // cannot infer an appropriate lifetime due to conflicting requirements
    E0566, // conflicting representation hints
    E0587, // conflicting packed and align representation hints
+    E0623, // lifetime mismatch where both parameters are anonymous regions


It'd be great if you could add a full description of the new error code, but that could of course be done in a subsequent PR.

will do it once I get in the anonymous region in Structs changes. Probably a separate PR

estebank · 2017-07-17T21:02:24Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+            let def_id = self.is_suitable_anonymous_region(region).unwrap();
+            let node_id = self.tcx.hir.as_local_node_id(def_id).unwrap();
+            let ret_ty = self.tcx.type_of(def_id);
+            match ret_ty.sty {


Given that in all of these you're matching only one possible element on the enum, you can also do

if let (ty::TyFnDef(..), hir_map::NodeItem(it)) = (ret_ty.sty, self.tcx.hir.get(node_id)) { if let hir::ItemFn(ref fndecl, ..) = it.node { fndecl .inputs ... }

estebank · 2017-07-17T21:03:55Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+                                            };
+                                            nested_visitor.visit_ty(&**arg);
+                                            if nested_visitor.found_type.is_some() {
+                                                return nested_visitor.found_type;


you can avoid the returns in both branches of the if (if you omit the ; as well) as the if is the last element of the closure.

estebank · 2017-07-17T21:05:44Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        } else {
+            None
+        }
+    }


Instead of all the } else { None } you can just return in the inner part of the chain (return fndecl.<...>.next()) and have a single None as the last line of the method.

Comment above.

if nested_visitor.found_type.is_some() { return nested_visitor.found_type; } }) .next() } _ => None, } } _ => None, } } _ => None, } } None }

gives me a missing else clause error.

The following should be (at least close to) correct and equivalent to the code you wrote:

fn find_anon_type(&self, region: Region<'tcx>, br: &ty::BoundRegion) -> Option<&hir::Ty> { if let Some(anon_reg) = self.is_suitable_anonymous_region(region) { let (def_id, _) = anon_reg; if let Some(node_id) = self.tcx.hir.as_local_node_id(def_id) { let ret_ty = self.tcx.type_of(def_id); if let ty::TyFnDef(_, _) = ret_ty.sty { if let hir_map::NodeItem(it) = self.tcx.hir.get(node_id) { if let hir::ItemFn(ref fndecl, _, _, _, _, _) = it.node { return fndecl .inputs .iter() .filter_map(|arg| { let mut nested_visitor = FindNestedTypeVisitor { infcx: &self, hir_map: &self.tcx.hir, bound_region: *br, found_type: None, }; nested_visitor.visit_ty(&**arg); if nested_visitor.found_type.is_some() { nested_visitor.found_type } else { None } }).next(); } } } } } None }

Notice the fall through for all the else conditions, with only one implicit return at the end.

estebank · 2017-07-17T21:22:51Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+            _ => return false, // inapplicable
+        };
+
+        // Determine whether the sub and sup consist of both anonymous (elided) regions.


This could be written a bit more succinctly:

// Determine whether the sub and sup consist of both anonymous (elided) regions. if let (Some(br1), Some(br2)) = (self.is_anonymous_region(sup), self.is_anonymous_region(sub)) { if let (Some(ty1), Some(t2)) = (self.find_anon_type(sup, &br1), self.find_anon_type(sub, &br2)) { if let (Some((anon_arg1, _, _, _)), Some((anon_arg2, _, _, _))) = (self.find_arg_with_anonymous_region(sup, sup), self.find_arg_with_anonymous_region(sub, sub)) { let span_label = if let (Some(var1), Some(var2)) = (anon_arg1.pat.simple_name(), anon_arg2.pat.simple_name()) { format!("data from `{}` flows into `{}` here", var1, var2)) } else { "data flows here".to_string() }; struct_span_err!(self.tcx.sess, span, E0623, "lifetime mismatch") .span_label(ty1.span, format!("these references must have the same lifetime")) .span_label(ty2.span, format!("")) err.span_label(span, span_label) .emit(); return true; } } } false

estebank · 2017-07-17T21:25:12Z

src/librustc/infer/error_reporting/mod.rs

-                                                     sup_origin, sup_r);
-                    }
-                }
+            if !self.try_report_named_anon_conflict(&error) && !self.try_report_anon_anon_conflict(&error){


(&error) {

estebank · 2017-07-17T21:25:30Z

src/librustc/infer/error_reporting/mod.rs

+                               sub_origin, sub_r,
+                               sup_origin, sup_r) => {
+                      self.report_sub_sup_conflict(var_origin,
+                                                 sub_origin, sub_r,


indentation missing two spaces

Fix indentation above.

arielb1 · 2017-07-17T21:48:04Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+            let span_label_var1 = if let Some(simple_name) = anon_arg1.pat.simple_name() {
+                format!("from `{}`", simple_name)
+            } else {
+                format!("data flows here")


This should be "" so the error would be "data flows into x here"

In the case of a complex pattern, the error would not show any of the parameters but just data flows here. Not sure if that's what you meant.

As the code is right now, in that case the label's text would be data data flows here flows here (also notice the double space between the last flows and here).

we should make tests for the four cases:

source and dest both have simple names

source has a simple name

destination has a simple name

no simple names

arielb1 · 2017-07-18T22:36:09Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+                match self.infcx.tcx.named_region_map.defs.get(&lifetime.id) {
+                    // the lifetime of the TyRptr!
+                    Some(&rl::Region::LateBoundAnon(debuijn_index, anon_index)) => {
+                        if debuijn_index.depth == 1 && anon_index == br_index {


nit: typo: should be debruijn.

arielb1 · 2017-07-18T22:38:47Z

I generally like the approach here, but we might also want to come up with a strategy for structs with anonymous lifetimes, and the code could be prettier.

gaurikholkar-zz · 2017-07-19T15:53:33Z


fn foo(z: &mut Vec<(&u8,&u8)>, (x, y): (&u8, &u8)) {
    z.push((x,y));
}

gives the error


error[E0623]: lifetime mismatch
  --> src/test/ui/lifetime-errors/ex3-both-anon-regions-2.rs:12:13
   |
11 | fn foo(z: &mut Vec<(&u8,&u8)>, (x, y): (&u8, &u8)) {
   |                     ---                 --- these references must have the same lifetime
12 |     z.push((x,y));
   |             ^ data  flows into `z` here

error[E0623]: lifetime mismatch
  --> src/test/ui/lifetime-errors/ex3-both-anon-regions-2.rs:12:15
   |
11 | fn foo(z: &mut Vec<(&u8,&u8)>, (x, y): (&u8, &u8)) {
   |                         ---                  --- these references must have the same lifetime
12 |     z.push((x,y));
   |               ^ data  flows into `z` here

error: aborting due to 2 previous errors

The report_anon_anon_conflict is getting called two times, once for z and x and then for z and y, treating x and y as two separate anonymous regions but the other testcases seem to be working fine.
Adding ui tests in a while.

@nikomatsakis I'm not sure whether the error message looks a bit too long.

gaurikholkar-zz · 2017-07-21T12:17:25Z

@arielb1 @estebank @nikomatsakis updated PR

estebank

Seems close, couple of nitpicks.

estebank · 2017-07-21T20:28:09Z

src/librustc/infer/error_reporting/mod.rs

-                    }
-                }
+            if !self.try_report_named_anon_conflict(&error) &&
+!self.try_report_anon_anon_conflict(&error) {


Fix indentation.

estebank · 2017-07-21T20:36:34Z

src/librustc/infer/error_reporting/util.rs

+
+                let id = free_region.scope;
+                let node_id = self.tcx.hir.as_local_node_id(id).unwrap();
+                let body_id = self.tcx.hir.maybe_body_owned_by(node_id).unwrap();


Avoid unwrap() in the compiler unless you are 110% sure that there's no way it will ever panic. On cases where you're 109% sure and below, either use ? to bubble up or use match/if let to handle the error with a bug!() message. If an error case is expected but doesn't need to be handled, just use if let to guard your new code against it. You can even do it with multiple variables at the same time:

let hir = self.tcx.hir; if let (Some(node_id), Some(body_id)) = (hir.maybe_body_owned_by(node_id), hir.as_local_node_id(id)) { ... }

estebank · 2017-07-21T20:37:08Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        } else {
+            None
+        }
+    }


Comment above.

nikomatsakis

I left some nits. Once @estebank is happy, I am also happy (i.e., r=me modulo the nits here and @estebank's approval).

(I see now that GH says it was showing me outdated commits; grr.... hopefully my comments are still relevant.)

With respect to the comment from @arielb1, the plan that @gaurikholkar and I had discussed was to handle cases like Foo<'a, ...> in a separate follow-up PR, and just focus here on the simpler &T case.

nikomatsakis · 2017-07-23T19:25:21Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+
+// The visitor captures the corresponding `hir::Ty` of the
+// anonymous region.
+struct FindNestedTypeVisitor<'a, 'gcx: 'a + 'tcx, 'tcx: 'a> {


It would be good to, in this comment, give an example of the error and show the role this visitor plays, I think.

e.g. "Consider a case where we have fn foo(x: &mut Vec<&u8>, y: &u8) { x.push(y); }. This would lead to a conflict between the two anonymous lifetimes for &u8 in x and y respectively. This visitor would be invoked twice, once for each lifetime, and would walk the types like &mut Vec<&u8> and &u8 looking for the HIR where that lifetime appears. This allows us to highlight the specific part of the type."

nikomatsakis · 2017-07-23T19:27:10Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+struct FindNestedTypeVisitor<'a, 'gcx: 'a + 'tcx, 'tcx: 'a> {
+    infcx: &'a InferCtxt<'a, 'gcx, 'tcx>,
+    hir_map: &'a hir::map::Map<'gcx>,
+    bound_region: ty::BoundRegion,


similarly, might be nice to add some (brief) comments to the fields, e.g. explaining that this is the bound-region we are looking for. These can reference your example snippet above.

nikomatsakis · 2017-07-23T19:27:21Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+    infcx: &'a InferCtxt<'a, 'gcx, 'tcx>,
+    hir_map: &'a hir::map::Map<'gcx>,
+    bound_region: ty::BoundRegion,
+    found_type: Option<&'gcx hir::Ty>,


and that this field (I presume) is an output of the visitor

nikomatsakis · 2017-07-23T19:28:15Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+    // This function calls the `visit_ty` method for the parameters
+    // corresponding to the anonymous regions. The `nested_visitor.found_type`
+    // contains the anonymous type.
+    pub fn find_anon_type(&self, region: Region<'tcx>, br: &ty::BoundRegion) -> Option<&hir::Ty> {


Nit: this comment seems to describe the implementation. I'd make a doc comment (using /// comments) and describe the interface to this function -- what does it take as input and return? What is it used for as part of the error reporting process?

The stuff in this comment can be moved inside the fn body, since the caller doesn't care about them.

Also, is this called from outside the file? If not, it need not be pub.

nikomatsakis · 2017-07-23T19:29:36Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+    // corresponding to the anonymous regions. The `nested_visitor.found_type`
+    // contains the anonymous type.
+    pub fn find_anon_type(&self, region: Region<'tcx>, br: &ty::BoundRegion) -> Option<&hir::Ty> {
+        if self.is_suitable_anonymous_region(region).is_some() {


Nit: Use if let Some(def_id) = self.is_suitable_anonymous_region(region) { or match instead of calling is_some() and then unwrap()

nikomatsakis · 2017-07-23T19:33:12Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        }
+    }
+
+    pub fn try_report_anon_anon_conflict(&self, error: &RegionResolutionError<'tcx>) -> bool {


It'd be great to have a comment explaining the kinds of errors this handles (give a concrete example). For that matter, I would try to move this function to the top of the file -- it's the "main entry point" for your reader. You can move the visitor struct etc, which are implementation details, down to the bottom of the file. Then you can also have the comments in those other stuff reference the example you introduce here.

nikomatsakis · 2017-07-23T19:33:43Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+                        .span_label(span,
+                                    format!("data{}flows{}here", span_label_var1, span_label_var2))
+                        .emit();
+                    return true;


I would also try to "un-nest" here, so that you have return false bailing early, and just focus on the main path.

nikomatsakis · 2017-07-23T19:33:55Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        false
+    }
+
+    pub fn is_anonymous_region(&self, region: Region<'tcx>) -> Option<ty::BoundRegion> {


Is this used from outside the file? If not, make it not be pub.

nikomatsakis · 2017-07-23T19:34:45Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+                        match self.tcx.hir.find(node_id) {
+                            Some(hir_map::NodeItem(..)) |
+                            Some(hir_map::NodeTraitItem(..)) => {
+                                // proceed ahead //


I would change this to something like: "// Success -- proceed to return Some below"

nikomatsakis · 2017-07-23T19:35:16Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+                                    // since the signature must match the trait.
+                                    //
+                                    // FIXME(#42706) -- in some cases, we could do better here.
+                                    return None;


In general, this function feels familiar -- I think we discussed this already, can it be combined with code from your other PR?

gaurikholkar-zz · 2017-07-26T22:24:33Z

@nikomatsakis @estebank updated PR

estebank

Almost there. I'm happy with the overall state of the PR, but have some minor comments still. As soon as you're done with the style changes, I'l r+.

estebank · 2017-07-26T22:32:23Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        }
+    }
+    /*
+    fn is_suitable_anonymous_region(&self, region: Region<'tcx>) -> Option<ty::BoundRegion> {


Remove commented out code.

estebank · 2017-07-26T22:38:05Z

src/librustc/infer/error_reporting/anon_anon_conflict.rs

+        } else {
+            None
+        }
+    }


The following should be (at least close to) correct and equivalent to the code you wrote:

fn find_anon_type(&self, region: Region<'tcx>, br: &ty::BoundRegion) -> Option<&hir::Ty> { if let Some(anon_reg) = self.is_suitable_anonymous_region(region) { let (def_id, _) = anon_reg; if let Some(node_id) = self.tcx.hir.as_local_node_id(def_id) { let ret_ty = self.tcx.type_of(def_id); if let ty::TyFnDef(_, _) = ret_ty.sty { if let hir_map::NodeItem(it) = self.tcx.hir.get(node_id) { if let hir::ItemFn(ref fndecl, _, _, _, _, _) = it.node { return fndecl .inputs .iter() .filter_map(|arg| { let mut nested_visitor = FindNestedTypeVisitor { infcx: &self, hir_map: &self.tcx.hir, bound_region: *br, found_type: None, }; nested_visitor.visit_ty(&**arg); if nested_visitor.found_type.is_some() { nested_visitor.found_type } else { None } }).next(); } } } } } None }

Notice the fall through for all the else conditions, with only one implicit return at the end.

estebank · 2017-07-26T22:38:34Z

src/librustc/infer/error_reporting/mod.rs

+                               sub_origin, sub_r,
+                               sup_origin, sup_r) => {
+                      self.report_sub_sup_conflict(var_origin,
+                                                 sub_origin, sub_r,


Fix indentation above.

estebank · 2017-07-26T22:40:28Z

src/librustc/infer/error_reporting/util.rs

+                                        region: Region<'tcx>)
+                                        -> Option<(DefId, ty::BoundRegion)> {
+        match *region {
+            ty::ReFree(ref free_region) => {


When matching on only one arm it is a good idea to use if let instead, as it communicates intent to not deal with the other possible matches, as well as yielding fewer lines of code by not having to add _ => ().

gaurikholkar-zz · 2017-07-27T19:14:58Z

@nikomatsakis updated

bors · 2017-07-27T22:22:38Z

☔ The latest upstream changes (presumably #43443) made this pull request unmergeable. Please resolve the merge conflicts.

estebank · 2017-07-28T02:27:36Z

Awesome! There's a merge conflict against master HEAD, but it shouldn't be too big of a problem. Do you know how to squash commits and rebase against latest master? I find it easier to rebase my work against master instead of merging, and rebasing one commit instead of multiple, which has the benefit of a slightly cleaner log.

Regardless, fix the merge conflict and I'll approve 👍

gaurikholkar-zz · 2017-07-28T02:56:41Z

@estebank updated.

estebank · 2017-07-28T03:12:03Z

@gaurikholkar two small tidy errors:

[00:03:44] tidy error: /checkout/src/librustc/infer/error_reporting/util.rs:87: trailing whitespace
[00:03:44] tidy error: /checkout/src/librustc/infer/error_reporting/util.rs:118: trailing whitespace

…mous

gaurikholkar-zz · 2017-07-28T05:37:35Z

@estebank done

estebank · 2017-07-28T15:46:46Z

@bors r+

bors · 2017-07-28T15:46:47Z

📌 Commit 4fb1808 has been approved by estebank

bors · 2017-07-28T16:03:39Z

⌛ Testing commit 4fb1808 with merge eba9d7f...

@nikomatsakis

improve case with both anonymous lifetime parameters #43269 This is a fix to #43269. Sample output message- ``` error[E0623]: lifetime mismatch --> $DIR/ex3-both-anon-regions.rs:12:12 | 11 | fn foo(x: &mut Vec<&u8>, y: &u8) { | --- --- these references must have the same lifetime 12 | x.push(y); | ^ data from `y` flows into `x` here error: aborting due to 2 previous errors ``` r? @nikomatsakis

bors · 2017-07-28T18:31:47Z

☀️ Test successful - status-appveyor, status-travis
Approved by: estebank
Pushing eba9d7f to master...

gaurikholkar-zz · 2017-07-29T12:16:23Z

#43541

rust-highfive assigned nikomatsakis Jul 17, 2017

estebank reviewed Jul 17, 2017

View reviewed changes

arielb1 reviewed Jul 17, 2017

View reviewed changes

arielb1 added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Jul 18, 2017

arielb1 reviewed Jul 18, 2017

View reviewed changes

gaurikholkar-zz force-pushed the lifetime_errors branch 2 times, most recently from bc177a9 to 9f82c94 Compare July 21, 2017 01:06

estebank reviewed Jul 21, 2017

View reviewed changes

gaurikholkar-zz force-pushed the lifetime_errors branch from 9f82c94 to b384a4a Compare July 23, 2017 19:24

nikomatsakis approved these changes Jul 23, 2017

View reviewed changes

estebank reviewed Jul 26, 2017

View reviewed changes

gaurikholkar-zz force-pushed the lifetime_errors branch from 3d4d2a4 to 2da6798 Compare July 28, 2017 02:55

Adding E0623, to detect missing lifetimes when both regions are anony…

4fb1808

…mous

gaurikholkar-zz force-pushed the lifetime_errors branch from 2da6798 to 4fb1808 Compare July 28, 2017 03:26

bors mentioned this pull request Jul 28, 2017

Implement tokenization for some items in proc_macro #43230

Merged

bors merged commit 4fb1808 into rust-lang:master Jul 28, 2017

bors mentioned this pull request Jul 28, 2017

Generator support #43076

Merged

7 tasks

		@@ -0,0 +1,201 @@
		// Copyright 2012-2013 The Rust Project Developers. See the COPYRIGHT

improve case with both anonymous lifetime parameters #43269 #43298

improve case with both anonymous lifetime parameters #43269 #43298

Conversation

gaurikholkar-zz commented Jul 17, 2017

estebank left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gaurikholkar-zz Jul 21, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

estebank Jul 17, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

estebank Jul 18, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

arielb1 commented Jul 18, 2017

gaurikholkar-zz commented Jul 19, 2017 • edited Loading

gaurikholkar-zz commented Jul 21, 2017

estebank left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nikomatsakis left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gaurikholkar-zz commented Jul 26, 2017

estebank left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gaurikholkar-zz commented Jul 27, 2017

bors commented Jul 27, 2017

estebank commented Jul 28, 2017

gaurikholkar-zz commented Jul 28, 2017

estebank commented Jul 28, 2017

gaurikholkar-zz commented Jul 28, 2017

estebank commented Jul 28, 2017

bors commented Jul 28, 2017

bors commented Jul 28, 2017

bors commented Jul 28, 2017

gaurikholkar-zz commented Jul 29, 2017

gaurikholkar-zz Jul 21, 2017 •

edited

Loading

estebank Jul 17, 2017 •

edited

Loading

estebank Jul 18, 2017 •

edited

Loading

gaurikholkar-zz commented Jul 19, 2017 •

edited

Loading