Document current coding style guidelines. #2298

Axel-Naumann · 2018-07-06T14:55:01Z

No description provided.

pcanal · 2018-07-06T20:48:32Z

README/RCodingStyle.hxx

+/// Functions that access only public members of ROOT classes shall be in here.
+/// \param a - first TExampleClass to add.
+/// \param b - second TExampleClass to add.
+double Add(const TExampleClass& a, const TExampleClass& b);


Did you mean to change the binding of the & (and *):
double Add(const TExampleClass &a, const TExampleClass &b);

yes please! :P

eguiraud

Nice, we really needed this! Is this retroactive for new classes? I.e., should I make RDataFrame & friends compliant?

eguiraud · 2018-07-06T19:50:58Z

README/RCodingStyle.cxx

@@ -0,0 +1,67 @@
+/// \file RCodingStyle.cxx
+/// \author Axel Naumann <axel@cern.ch>


how does this look for multiple authors?

Should be addressed.

eguiraud · 2018-07-06T19:51:14Z

README/RCodingStyle.cxx

@@ -0,0 +1,67 @@
+/// \file RCodingStyle.cxx
+/// \author Axel Naumann <axel@cern.ch>
+/// \date 2018-07-06


what date is this? creation or last modification?

Should be addressed.

eguiraud · 2018-07-06T19:52:30Z

README/RCodingStyle.cxx

+
+// This file demonstrates the coding styles to be used for new ROOT files.
+// For much of the rationale see Scott Meyer's "Efficient Modern C++" and the
+// [C++ core guildlines](https://github.com/isocpp/CppCoreGuidelines/blob/master/CppCoreGuidelines.md),


"guildlines"

eguiraud · 2018-07-06T19:52:48Z

README/RCodingStyle.cxx

+ * For the list of contributors see $ROOTSYS/README/CREDITS.             *
+ *************************************************************************/
+
+// This file demonstrates the coding styles to be used for new ROOT files.


"style" singular?

eguiraud · 2018-07-06T20:14:07Z

README/RCodingStyle.cxx

+// Now the same order as for the header:
+// ROOT headers in alphabetical order; all new ROOT headers are under `ROOT/`
+#include "ROOT/TSeq.hxx"
+// Old ones are found in include/ without prefix. Try to use the new ones.


what do you mean with "Try to use the new ones"? The percentage of functionality that is available both in ROOT/R* headers and in T* headers is very little afaik?

Or is this a "Try out the new ones when available!" (as in, advertisement more than guideline)? In which case my comment would be that there is no way for users to know whether ROOT offers a ROOT/RString.hxx as a modern substitute for TString.h, except for searching for it in the doxygen docs case by case. And even then the names are not always the same, e.g. the modern TH1F.h is not ROOT/RH1F.hxx.

eguiraud · 2018-07-06T21:03:42Z

README/RCodingStyle.hxx

+/// Disabled if `T` is of reference type.
+
+template <int IDX, // Template parameters are all-capital letters
+          class T, // We use "class", not "typename"; `T` is fine for generic class names


damn, that's a lot of search-and-replace in RDataFrame's headers 😋

No worries. Some inconsistencies are fine; people won't even notice. It's more important that we agree on what to do from here on.

eguiraud · 2018-07-06T21:19:07Z

README/RCodingStyle.hxx

+/// on the template parameters, e.g. `IDX` must be positive, `T` must be  copyable.
+/// Disabled if `T` is of reference type.
+
+template <int IDX, // Template parameters are all-capital letters


this is the only new convention that I don't like, it makes template-heavy code a bit heavy to look at (just go to TypeTraits.hxx and capitalize all template parameters...) and macro-like.

It helps to identify template parameters as such. I often use a using declaration to "convert" the template param to a type (with type-spelling).

But I see your point. Let's discuss at a team meeting!

eguiraud · 2018-07-06T21:25:37Z

README/RCodingStyle.hxx

+
+   // Use `= delete` instead of private, unimplemented.
+   // To avoid copy and move construction and assignment, `delete` the copy
+   // constructor and copy-assignment operator:


does this mean that you prefer to not have the move-ctor deleted?

eguiraud · 2018-07-06T21:30:19Z

README/RCodingStyle.hxx

+   /// they "inherit" the virtuality of the virtual base class destructor).
+   virtual ~RCodingStyle() {}
+
+   /// Overridden virtual functions do not specify `virtual` but `override` or `final`.


I would say either override or final: exactly one of the two, as per CppCoreGuidelines

eguiraud · 2018-07-06T21:32:14Z

README/RCodingStyle.hxx

+/// Functions that access only public members of ROOT classes shall be in here.
+/// \param a - first TExampleClass to add.
+/// \param b - second TExampleClass to add.
+double Add(const TExampleClass& a, const TExampleClass& b);


RExampleClass

Axel-Naumann · 2018-07-09T09:29:28Z

I don't think we should invest effort in retrofitting this to RDataFrame, but you're welcome to revisit parts as you see fit. Let's first agree on the style. I will present this in a team meeting to discuss it; e.g. the template arguments will be part of this discussion.

amadio · 2018-07-06T16:24:30Z

README/RCodingStyle.cxx

+static const std::string &StaticFunc() {
+   // NOTE that this initialization is thread-safe!
+   static std::string sString = "SomeString";
+   return sString;


What about return "SomeString";?

a) that's not the point of this demo-function, and b) that won't give you a string&.

I have improved the code, I hope it's less confusing now?

Yes, but maybe git commit --amend is better than adding a separate change for this.

amadio · 2018-07-06T16:34:28Z

README/RCodingStyle.hxx

+ *                                                                       *
+ * For the licensing terms see $ROOTSYS/LICENSE.                         *
+ * For the list of contributors see $ROOTSYS/README/CREDITS.             *
+ *************************************************************************/


Do we need this header on every file? The top 3 lines are obsolete almost immediately, and git is there for that, while the copyright can be in a file at top level stating that that's the copyright unless a different one exists in the file. Also, many files don't have the header in any case:

epsft-53 root $ git ls-files | grep '\.[ch][x][x]' | wc -l 2847 epsft-53 root $ git ls-files | grep '\.[ch][x][x]' | xargs -n1 grep 'Rene Brun' | wc -l 1739

and many others are out of date:

epsft-53 root $ grep -hor '1995-[0-9]\+' | sort | uniq -c 4 1995-01 1 1995-10 4 1995-1996 114 1995-1997 3 1995-1998 3 1995-1999 1308 1995-2000 70 1995-2001 149 1995-2002 49 1995-2003 279 1995-2004 213 1995-2005 137 1995-2006 321 1995-2007 184 1995-2008 48 1995-2009 20 1995-2010 55 1995-2011 57 1995-2012 77 1995-2013 11 1995-2014 44 1995-2015 53 1995-2016 89 1995-2017 29 1995-2018 1 1995-8

The top 3 lines are obsolete almost immediately, and git is there for that,

There is a difference between touching a file and writing it. I have argued before that this is a nice way of leaving traces also for people with limited participation in ROOT, and it's a good way of knowing whom to contact / ownership.

the copyright can be in a file at top level stating that that's the copyright unless a different one exists in the file

Multiple opinions exist on that one.

Also, many files don't have the header in any case

Indeed we have many files that have a different (c).

There is a difference between touching a file and writing it. I have argued before that this is a nice way of leaving traces also for people with limited participation in ROOT, and it's a good way of knowing whom to contact / ownership.

I still think that git blame is more than sufficient for that, but the final choice is up to you.

amadio · 2018-07-06T16:36:46Z

README/RCodingStyle.hxx

+// The first non-comment, non-empty line must be the #include guard.
+// Acceptable forms are:
+//   ROOT_RCodingStyle, ROOT_RCodingStyle_hxx, ROOT_RCodingStyle_Hxx,
+//   ROOT_RCODINGSTYLE, ROOT_RCODINGSTYLE_HXX


Why not choose one only?

What or whom could a difference here confuse? What's the advantage of choosing one? They should be file-local, I sort of like that nobody can rely on them outside this file ;-)

We should likely mangle at least the component path in include protector. It may become very ugly to debug if the include protector names are duplicate between two unrelated files.

I'd make sense to me to stick with one include protector style (the latter).

Choosing only one is good for consistency. Isn't this trying to improve exactly that? I don't really care which one we choose, as long as it's consistent throughout the code. At the very least one choice should be recommended over the others. There's also #pragma once, but since it's not in any standard, maybe we should stay away from that.

vgvassilev · 2018-07-17T15:32:06Z

README/RCodingStyle.hxx

+
+// Especially within header files, reduce the number of included headers as
+// much as possible, at the expense of more forward declarations.
+// EXCEPTION: standard library classes must be #included, not forward declared.


Even the easy-to-forward-declare ones?

Doh, I get your point :)

vgvassilev · 2018-07-17T15:35:17Z

README/RCodingStyle.hxx

+// Include non-ROOT, non-stdlib headers last.
+// Rationale: non-ROOT, non-stdlib headers often #define like mad. Reduce interference
+// with ROOT or stdlib headers.
+#include <png.h>


That sounds weird. In principle png.h may depend on stl. That's not a problem for non-ROOT packages but what about the ones that are on the borderline: eg libafterimage and others that are patched by us.

png.h may depend on stl

No. System headers never depend on C++ and should thus come last.

vgvassilev · 2018-07-17T15:36:22Z

README/RCodingStyle.hxx

+   static std::string fgNameBadPleaseUseStaticFunctionStaticVariable;
+
+protected:
+   /// FOURTH section: non-public functions.


There seems to be no THIRD.

// THIRD section: data members.?

vgvassilev · 2018-07-17T15:39:00Z

README/RCodingStyle.hxx

+double Add(const TExampleClass &a, const TExampleClass &b);
+
+} // namespace ROOT
+#endif // #ifndef ROOT_RCodingStyle


I'd omit the #ifndef, it's meaning is implied by the previous token ;)

phsft-bot · 2018-07-23T08:53:27Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

phsft-bot · 2018-07-23T08:59:33Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

amadio · 2018-07-23T09:20:06Z

README/RCodingStyle.hxx

@@ -174,5 +174,5 @@ public:
 double Add(const TExampleClass &a, const TExampleClass &b);

 } // namespace ROOT
-#endif // #ifndef ROOT_RCodingStyle
+#endif // ROOT_RCodingStyle


Commenting on the commit message... Maybe a Suggested-by: tag in the body of the commit would be better?
There's a JIRA issue on me for the style of commit messages, maybe we should discuss this today too?

phsft-bot · 2018-07-23T11:01:59Z

Build failed on slc6/gcc48.
See console output.

Errors:

ERROR: Timeout after 30 minutes
ERROR: Error fetching remote repo 'origin'
ERROR: Error fetching remote repo 'origin'
ERROR: Timeout after 30 minutes
ERROR: Error fetching remote repo 'origin'
ERROR: Error fetching remote repo 'origin'

phsft-bot · 2018-07-23T12:46:20Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

phsft-bot · 2018-07-23T13:51:14Z

Build failed on slc6/gcc48.
See console output.

Errors:

ERROR: Timeout after 30 minutes
ERROR: Error cloning remote repo 'origin'
ERROR: Error cloning remote repo 'origin'

phsft-bot · 2018-07-25T10:56:22Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

phsft-bot · 2018-07-25T10:57:26Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

phsft-bot · 2018-07-25T10:59:47Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dimt=ON -Dccache=ON
How to customize builds

phsft-bot · 2018-07-25T12:10:39Z

Build failed on centos7/clang39.
See console output.

Errors:

ERROR: Timeout after 30 minutes
ERROR: Error cloning remote repo 'origin'
ERROR: Error cloning remote repo 'origin'

vgvassilev · 2018-07-27T07:58:30Z

README/RCodingStyle.cxx

@@ -0,0 +1,83 @@
+/// \file RCodingStyle.cxx


Sometimes it makes sense to add \brief section which describes the contents and the purpose of the files.

eguiraud · 2018-08-07T08:56:58Z

README/RCodingStyle.hxx

+// Parts of ROOT might want to use sub-namespaces for *internal* code.
+// Such code might include the following two namespaces:
+namespace Detail {
+// Contains code that is not expected to be seen by the everage user.


typo everage

phsft-bot · 2018-08-27T16:42:42Z

Starting build on slc6/gcc48, slc6-i686/gcc49, centos7/clang39, centos7/gcc62, centos7/gcc7, fedora28/native, ubuntu16/native, mac1013/native, windows10/vc15 with flags -Dccache=ON
How to customize builds

bernhardmgruber · 2020-07-29T12:48:11Z

README/RCodingStyle.cxx

+ROOT::RExampleClass::~RExampleClass()
+{
+}


why not ROOT::RExampleClass::~RExampleClass() = default; ?

bernhardmgruber · 2020-07-29T12:53:04Z

README/RCodingStyle.hxx

+// Include ROOT's headers first, in alphabetical order:
+//   Rationale: make sure that ROOT headers are standalone.
+// Do not use #include guards at the point of `#include`.
+//   Rationale: compilers optimize this themselves, this is interfering with their optimization.
+// Use <>, not "".
+//   Rationale: "" looks at ./ first; as ROOT headers are included by user code
+//   that adds context fragility.
+#include <ROOT/TSeq.hxx>
+// Include v6 ROOT headers last.
+// You're welcome to state what's used from a header, i.e. why it's included.
+#include <Rtypes.h> // for ULong64_t
+
+// Now include other C++ headers hat are not from the stdlib:
+// e.g. #include <cppyy/CPpYy.h>
+
+// Include stdlib headers next, in alphabetical order.
+#include <string>
+
+// Include C headers last.
+// Rationale: non-ROOT, non-stdlib headers often #define like mad. Reduce interference
+// with ROOT or stdlib headers.
+#include <curses.h>


It would be great if these rules could be automated via the IncludeCategories setting in clang-format.

bernhardmgruber · 2020-07-29T12:56:34Z

README/RCodingStyle.hxx

+   /// Static functions line up alphabetically with the others.
+   static int GetSeq() { return kSeq; }
+
+   /// It is fine to have one-line function declarations.


I guess you mean function definitions.

bernhardmgruber · 2020-07-29T12:57:55Z

README/RCodingStyle.hxx

+   /// \param[in] input A collection of strings to use.
+   /// \return Always return true but if it could fail would return false in that case.
+   ///
+   /// All functions must have at least one line of comments, parseable/useable by doxygen.


That sounds very strict to me. Simple functions with good names should not need comments.

I agree and disagree. One line of comment should be mandatory, because otherwise the docs look like this:
https://root.cern/doc/master/classTInterpreter.html

This one line should be one sentence, ending with a ., so doxygen automatically recognises this as a \brief.

I agree, though, that after having written a single sentence, you may leave out the rest for trivial functions.

bernhardmgruber · 2020-07-29T13:04:18Z

README/RCodingStyle.hxx

+
+   /// Virtual destructors of derived classes may be decorated with `virtual` (even though
+   /// they "inherit" the virtuality of the virtual base class destructor).
+   virtual ~RCodingStyle() {}


Should probably be virtual ~RCodingStyle() = default;.

hageboeck · 2020-07-30T15:10:22Z

README/RCodingStyle.hxx

+   /// \param[in] input A collection of strings to use.
+   /// \return Always return true but if it could fail would return false in that case.
+   ///
+   /// All functions must have at least one line of comments, parseable/useable by doxygen.


I agree and disagree. One line of comment should be mandatory, because otherwise the docs look like this:
https://root.cern/doc/master/classTInterpreter.html

This one line should be one sentence, ending with a ., so doxygen automatically recognises this as a \brief.

I agree, though, that after having written a single sentence, you may leave out the rest for trivial functions.

hageboeck · 2020-07-30T15:15:45Z

README/RCodingStyle.hxx

+//   that adds context fragility.
+#include <ROOT/TSeq.hxx>
+// Include v6 ROOT headers last.
+// You're welcome to state what's used from a header, i.e. why it's included.


In practice, this might be outdated the second time a file is touched.

Fully agree!

hageboeck · 2020-07-30T15:24:13Z

README/RCodingStyle.cxx

+}
+
+/// Instead of static data members (whether private or public), use outlined static
+/// functions. This solves the static initializion fiasco, and delays the initialization


Suggested change

/// functions. This solves the static initializion fiasco, and delays the initialization

/// functions. This solves the static initialization fiasco, and delays the initialization

hageboeck · 2020-07-30T15:33:01Z

README/RCodingStyle.hxx

+};
+
+///\class RCodingStyle
+/// This is exhaustive documentation of the class template, explaining the constraints


The first thing after the \class macro should be the brief documentation. Doxygen takes the first sentence of the docs if no \brief is passed, so try to just write one sentence in one line that ends with a ..

hageboeck · 2020-07-30T15:35:42Z

README/RCodingStyle.hxx

+///\class RCodingStyle
+/// This is exhaustive documentation of the class template, explaining the constraints
+/// on the template parameters, e.g.
+///   `IDX` must be positive, `T` must be copyable.


Recommend using doxygen's\tparam, e.g.:

Suggested change

/// `IDX` must be positive, `T` must be copyable.

/// \tparam IDX The index type. Must be positive.

/// \tparam T A class that is copyable.

hageboeck · 2020-07-30T15:39:50Z

README/RCodingStyle.hxx

+   RCodingStyle() = default;
+
+   // Open a group:
+   /// \{ \name copy pperations


Suggested change

/// \{ \name copy pperations

/// \{ \name copy operations

hageboeck · 2020-07-30T15:42:31Z

README/RCodingStyle.hxx

+// Parts of ROOT might want to use sub-namespaces for *internal* code.
+// Such code might include the following two namespaces:
+namespace Detail {
+// Contains code that is not expected to be seen by the everage user.


Suggested change

// Contains code that is not expected to be seen by the everage user.

// Contains code that is not expected to be seen by the average user.

hageboeck · 2020-07-30T15:43:05Z

README/RCodingStyle.hxx

+/// \param a - first RExampleClass to add.
+/// \param b - second RExampleClass to add.


I think that doxygen generates better docs without the -:

Suggested change

/// \param a - first RExampleClass to add.

/// \param b - second RExampleClass to add.

/// \param a First RExampleClass to add.

/// \param b Second RExampleClass to add.

Axel-Naumann requested review from pcanal, dpiparo and eguiraud July 6, 2018 14:55

Axel-Naumann self-assigned this Jul 6, 2018

Axel-Naumann requested a review from amadio July 6, 2018 16:10

pcanal reviewed Jul 6, 2018

View reviewed changes

eguiraud reviewed Jul 6, 2018

View reviewed changes

Axel-Naumann force-pushed the CodingStyle branch 3 times, most recently from 0b18907 to f78f9bd Compare July 9, 2018 09:25

root-project deleted a comment from phsft-bot Jul 9, 2018

amadio reviewed Jul 9, 2018

View reviewed changes

vgvassilev reviewed Jul 17, 2018

View reviewed changes

amadio reviewed Jul 23, 2018

View reviewed changes

Axel-Naumann added 5 commits July 23, 2018 13:56

Document current coding style guidelines.

1366385

[CodingStyle] Clarify the role of date, author.

5922dd4

[CodingStyle] Add rationale for delete lvalue ctor/assign.

7fae453

[CodingStyle] Remove "#ifndef" from #endif comment. Thanks, Vassil!

9cc9105

[CodingStyle] Clarify static init; show "why we include" comment.

31ec4ab

Axel-Naumann added 4 commits July 23, 2018 14:05

[CodingStyle] typo.

ea2bfbd

[CodingStyle] Fix C++, show fwd decl of ROOT class.

818f111

[CodingStyle] Need to #include Rtypes.h, mention <>, formatting.

a8a6b89

[CodingStyle] curses.h is more common than png.h.

68471fc

Axel-Naumann force-pushed the CodingStyle branch from dcc01dc to 68471fc Compare July 23, 2018 12:45

Axel-Naumann added 3 commits July 25, 2018 08:55

FIXUP - team review comments.

89aafac

[CodingStyle] Remove base splicing example: too specific.

9d7e799

[CodingStyle] Apply team review comments.

3c9c6cb

[codingStyle] Allow "July, 2018" as \date.

c97efc0

FIXUP

0d1fbce

vgvassilev reviewed Jul 27, 2018

View reviewed changes

eguiraud reviewed Aug 7, 2018

View reviewed changes

eguiraud mentioned this pull request Aug 27, 2018

SQlite data source for RDataFrame #2322

Merged

Mention the need for documentation.

79b51a9

eguiraud added the stale label Mar 10, 2020

Axel-Naumann mentioned this pull request Jul 29, 2020

[WIP] Applying modernize-use-override #6007

Closed

bernhardmgruber reviewed Jul 29, 2020

View reviewed changes

hageboeck reviewed Jul 30, 2020

View reviewed changes

Axel-Naumann removed the stale label Sep 14, 2020

ferdymercury added the in:Documentation label Apr 15, 2024

ferdymercury unassigned Axel-Naumann Apr 26, 2024

couet assigned dpiparo Sep 9, 2024

		@@ -0,0 +1,67 @@
		/// \file RCodingStyle.cxx
		/// \author Axel Naumann <axel@cern.ch>

	/// functions. This solves the static initializion fiasco, and delays the initialization
	/// functions. This solves the static initialization fiasco, and delays the initialization

	/// `IDX` must be positive, `T` must be copyable.
	/// \tparam IDX The index type. Must be positive.
	/// \tparam T A class that is copyable.

	// Contains code that is not expected to be seen by the everage user.
	// Contains code that is not expected to be seen by the average user.

		/// \param a - first RExampleClass to add.
		/// \param b - second RExampleClass to add.

Document current coding style guidelines. #2298

Are you sure you want to change the base?

Document current coding style guidelines. #2298

Uh oh!

Conversation

Axel-Naumann commented Jul 6, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

eguiraud left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Axel-Naumann Jul 9, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Axel-Naumann commented Jul 9, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vgvassilev Jul 17, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Axel-Naumann Jul 9, 2018 •

edited

Loading

vgvassilev Jul 17, 2018 •

edited

Loading

amadio Jul 23, 2018 •

edited

Loading