Misc warnings by seanm · Pull Request #2801 · fmtlib/fmt

seanm · 2022-03-07T19:19:09Z

No description provided.

vitaut

Thanks for the PR.

vitaut · 2022-03-07T19:57:20Z

  using pointer = void;
  using reference = void;
-  using _Unchecked_type =
+  using Unchecked_type_ =


This will introduce warnings in MSVC, please revert.

What warning would it introduce?

Underscore followed by uppercase is reserved in C++: https://en.cppreference.com/w/cpp/language/identifiers

Warnings about (un)checked iterators.

Is _Unchecked_type some kind of Microsoft extension keyword?

Sort of. It's not a keyword but some mechanism to opt out iterators.

Could we somewhere do something like:

#if defined(_MSC_VER) #define fmt_unchecked_type _Unchecked_type #else #define fmt_unchecked_type #endif

If you have access to MSVC 2015, I'd recommend checking if removing _Unchecked_type altogether still produces warnings. It might be no longer the case as IIRC they have been moving away from the checked iterators. You could even submit a test PR to the fmt repo and check warnings in appveyor: https://ci.appveyor.com/project/vitaut/fmt.

Actually, we can see on this PR that there are definitely more warnings: https://ci.appveyor.com/project/vitaut/fmt/builds/42816914/job/briwktqjiiw9oa34. The macro approach seems OK but note that macros should be uppercase.

OK, see new attempt.

vitaut · 2022-03-07T19:57:29Z

  using std::back_insert_iterator<detail::buffer<char>>::back_insert_iterator;
  appender(base it) noexcept : base(it) {}
-  using _Unchecked_type = appender;  // Mark iterator as checked.
+  using Unchecked_type_ = appender;  // Mark iterator as checked.


vitaut · 2022-03-09T19:03:25Z

 #ifndef FMT_COMPILE_H_
 #define FMT_COMPILE_H_

+#include "core.h"


This is not needed: core.h is a lightweight alternative to format.h, it doesn't make sense to include it when format.h is included.

vitaut · 2022-03-09T19:07:17Z

+#ifdef _MSC_VER
+#  define FMT_UNCHECKED_TYPE _Unchecked_type
+#else
+#  define FMT_UNCHECKED_TYPE DummyName
+#endif


Let's change it into:

#define FMT_UNCHECKED_ITERATOR(It) using _Unchecked_type = It; // Mark iterator as checked.

to simplify call sites.

vitaut · 2022-03-09T19:07:57Z

 // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61414.
 namespace align {
-enum type { none, left, right, center, numeric };
+enum type : unsigned { none, left, right, center, numeric };


Why unsigned?

As per commit message: to fix the -Wsigned-enum-bitfield warnings.

vitaut · 2022-03-10T20:13:44Z

+  using FMT_UNCHECKED_TYPE =
      truncating_iterator_base;  // Mark iterator as checked.


This should be `FMT_UNCHECKED_TYPE(truncating_iterator_base);

Doh, sorry I missed that one. Fixed.

vitaut · 2022-03-10T20:14:40Z

+#  define FMT_UNCHECKED_ITERATOR(It) using _Unchecked_type = It; // Mark iterator as checked.
+#else
+#  define FMT_UNCHECKED_ITERATOR(It) using DummyTypeName = It;


Let's remove trailing semicolons since those are added at call sites.

Ah, good point. Fixed.

vitaut · 2022-03-11T16:20:34Z

Please fix build failures.

vitaut · 2022-03-11T17:24:07Z

And apply clang-format.

seanm · 2022-03-15T18:42:13Z

Please fix build failures.

I don't understand that error. I played on godbolt, and it seems to me like a gcc bug fixed between 9.2 and 9.3:

https://godbolt.org/z/58aEv8zEq

Interesting, the warning goes away if I change the enum from unsigned to unsigned char. Shall we do that?

vitaut · 2022-03-15T19:19:11Z

Interesting, the warning goes away if I change the enum from unsigned to unsigned char. Shall we do that?

If that resolves the issue then sure.

vitaut · 2022-03-17T17:25:23Z

There are still build failures.

vitaut · 2022-03-24T04:05:24Z

Do you still plan working on this PR?

seanm · 2022-03-24T13:41:53Z

Do you still plan working on this PR?

Yes. But I'm at a loss as to why that old gcc complains. I don't have a means to reproduce it locally.

I guess I could split that one commit out so the rest can be merged...

Avoid defining various reserved identifiers (starting with underscore and capital letter). Fortunately, they were all Windows-only, so it was easy to conditionalize them in Window-only preprocessor checks.

Created FMT_UNCHECKED_TYPE that resolves to special identifier _Unchecked_type for Microsoft, but to a dummy string otherwise. Using _Unchecked_type is invalid because underscore + uppercase is a reserved identifier.

vitaut · 2022-03-24T14:33:47Z

Thanks

vitaut requested changes Mar 7, 2022

View reviewed changes

phprus reviewed Mar 8, 2022

View reviewed changes

Comment thread test/test-main.cc

seanm force-pushed the misc-warnings branch from 69430ff to b30e335 Compare March 9, 2022 18:28

vitaut requested changes Mar 9, 2022

View reviewed changes

seanm force-pushed the misc-warnings branch from b30e335 to b9560da Compare March 9, 2022 19:58

vitaut requested changes Mar 10, 2022

View reviewed changes

seanm force-pushed the misc-warnings branch from b9560da to bce2c3b Compare March 10, 2022 20:37

seanm force-pushed the misc-warnings branch from bce2c3b to 9adbbcd Compare March 15, 2022 19:38

seanm added 2 commits March 24, 2022 09:57

Fixed all clang -Wreserved-id-macro warnings (on macOS at least)

c6d7691

Avoid defining various reserved identifiers (starting with underscore and capital letter). Fortunately, they were all Windows-only, so it was easy to conditionalize them in Window-only preprocessor checks.

Fixed clang -Wreserved-identifier warings

098b9d2

Created FMT_UNCHECKED_TYPE that resolves to special identifier _Unchecked_type for Microsoft, but to a dummy string otherwise. Using _Unchecked_type is invalid because underscore + uppercase is a reserved identifier.

seanm force-pushed the misc-warnings branch from 9adbbcd to 098b9d2 Compare March 24, 2022 13:58

vitaut approved these changes Mar 24, 2022

View reviewed changes

vitaut merged commit 5379063 into fmtlib:master Mar 24, 2022

seanm mentioned this pull request May 8, 2022

Fixed all clang -Wsigned-enum-bitfield warnings #2882

Merged

		using FMT_UNCHECKED_TYPE =
		truncating_iterator_base; // Mark iterator as checked.

Uh oh!

Conversation

seanm commented Mar 7, 2022

Uh oh!

vitaut 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!

vitaut Mar 8, 2022 • 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!

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!

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!

Uh oh!

vitaut commented Mar 11, 2022

Uh oh!

vitaut commented Mar 11, 2022

Uh oh!

seanm commented Mar 15, 2022

Uh oh!

vitaut commented Mar 15, 2022

Uh oh!

vitaut commented Mar 17, 2022

Uh oh!

vitaut commented Mar 24, 2022

Uh oh!

seanm commented Mar 24, 2022

Uh oh!

vitaut commented Mar 24, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

vitaut Mar 8, 2022 •

edited

Loading