Fix build problems w/ fmt and C++23 #1102

ltjax · 2025-12-30T16:18:44Z

PR for Issue #1101
Pretty superficial fixes, e.g. formatOptionalFileOffset could probably be refactored to work without fmt::runtime, but this was less invasive.

Unfortunately, it doesn't compile with C++ 20 and 23 for me anymore after I rebased onto main, but due to different problems, i.e.:

/home/ltjax/code/KTX-Software/tools/ktx/command_convert.cpp:208:65: error: no matching function for call to ‘ktx::OutputStreamEx::OutputStreamEx(std::u8string, ktx::CommandConvert&)’
  208 |     OutputStreamEx outputStream(outputFilepath.u8string(), *this);
      |                                                                 ^
/home/ltjax/code/KTX-Software/tools/ktx/command_convert.cpp:82:5: note: candidate: ‘ktx::OutputStreamEx::OutputStreamEx(const std::string&, ktx::Reporter&)’
   82 |     OutputStreamEx(const std::string& filepath, Reporter& report)
      |     ^~~~~~~~~~~~~~
/home/ltjax/code/KTX-Software/tools/ktx/command_convert.cpp:82:39: note:   no known conversion for argument 1 from ‘std::u8string’ {aka ‘std::__cxx11::basic_string<char8_t>’} to ‘const std::string&’ {aka ‘const std::__cxx11::basic_string<char>&’}
   82 |     OutputStreamEx(const std::string& filepath, Reporter& report)
      |                    ~~~~~~~~~~~~~~~~~~~^~~~~~~~

I don't know how to proceed there. You seem to have put some thought into utf8 handling and might better know how to fix that one.

CLAassistant · 2025-12-30T16:18:50Z

All committers have signed the CLA.

MarkCallow

I have only the one question.

MarkCallow · 2026-01-01T07:25:58Z

tools/ktx/command_compare.cpp

    virtual std::string value(std::size_t index, OutputFormat format) const override {
        if (format == OutputFormat::text) {
-            return fmt::format(fmt::format("0x{{:0{}x}}", sizeof(T) << 1), DiffBase<T>::rawValue(index));
+            return fmt::format(fmt::runtime(fmt::format("0x{{:0{}x}}", sizeof(T) << 1)), DiffBase<T>::rawValue(index));


What is the fmt::runtime for?

It just turns off the compile-time checking of the format string for the outer format call. I basically added it whenever the format string was runtime-generated.

Actually, I think in this case, you can probably get away with just a single format call and get rid of fmt::runtime. AFAIK, fmt can deal with placeholders in placeholders just fine.

I'd prefer if we can get rid of it.

MarkCallow · 2026-01-02T06:50:03Z

/home/ltjax/code/KTX-Software/tools/ktx/command_convert.cpp:208:65: error: no matching function for call to > ‘ktx::OutputStreamEx::OutputStreamEx(std::u8string, ktx::CommandConvert&)’
 208 |     OutputStreamEx outputStream(outputFilepath.u8string(), *this);
     |                                                                 ^

Sorry. I did not pay proper attention to your description. I initially thought this was a problem this PR fixes.

The issue is that c++20 supports the new u8string type and filesystem::path::u8string() returns a u8string instead of a string. The fix is to add an overload for the OutputStreamEx constructor that takes a u8string. Implementation may require u8string-supporting versions of our utf8 file handling utilities. I have just created those, but they are in a branch I'm working on. I'll make a PR as soon as I can with those, some fixes I am already making in command.cpp where OutputStreamEx is defined, and a fix for this issue. I'll ask you to rebase when that is done. In the meantime please answer the question in my review.

MarkCallow · 2026-01-03T08:17:45Z

The fix for the OutputStreamEx constructor for c++20 is now in main from PR #1103. Please rebase this.

ltjax · 2026-01-03T15:17:07Z

Ok, I rebased and removed that fmt::runtime, but it still does not compile with C++20 and C++23, albeit with different errors:

For CMAKE_CXX_STANDARD=23 I'm getting:

/home/ltjax/code/KTX-Software/external/basis_universal/encoder/cppspmd_sse.h:496:24: error: cannot bind non-const lvalue reference of type ‘cppspmd_sse41::spmd_kernel::vfloat&’ to an rvalue of type ‘cppspmd_sse41::spmd_kernel::vfloat’
  496 |                 return dst;
      |                        ^~~
/home/ltjax/code/KTX-Software/external/basis_universal/encoder/cppspmd_sse.h: In member function ‘cppspmd_sse41::spmd_kernel::vfloat& cppspmd_sse41::spmd_kernel::store_all(vfloat&&, const vfloat&)’:
/home/ltjax/code/KTX-Software/external/basis_universal/encoder/cppspmd_sse.h:508:24: error: cannot bind non-const lvalue reference of type ‘cppspmd_sse41::spmd_kernel::vfloat&’ to an rvalue of type ‘cppspmd_sse41::spmd_kernel::vfloat’
  508 |                 return dst;
      |

Not sure what the intention is with those, since the type does not really seem movable? But it's an external anyways, so I'm not sure whether you want that fixed.

For CMAKE_CXX_STANDARD=20, the first error I'm getting is:

/home/ltjax/code/KTX-Software/utils/platform_utils.h:125:21: error: could not convert ‘std::__cxx11::basic_string<char8_t>((& s)->std::__cxx11::basic_string<char>::begin(), (& s)->std::__cxx11::basic_string<char>::end(), std::allocator<char8_t>())’ from ‘basic_string<char8_t>’ to ‘basic_string<char>’
  125 |         return std::u8string(s.begin(), s.end());
      |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                     |
      |                     basic_string<char8_t>

I think the return type should be std::u8string, but even after fixing that there are still a lot of utf8 related errors, e.g. the additional OutputStream (without Ex) constructor has OutputStream::, but it defined in the class. And InputStream and fmtInFile in command_compare still cannot deal with std::u8string.

MarkCallow · 2026-01-04T07:32:24Z

I think the return type should be std::u8string

Sorry for the stupid copy and paste error. The return type should indeed be std::u8string. Please fix it in this PR.

e.g. the additional OutputStream (without Ex) constructor has OutputStream::, but it defined in the class.

Sorry about that too. I started implementation in command.cpp but when I realized it would simply convert the input and call the other constructor I moved it to the class definition.

I do not understand how everything built in our CI. All of the libktx tests use c++20 and two of them include platform_utils.h.Most peculiar.

I have fixed the InputStream and fmtInFile issues. Please see the c++fixes branch for the fixes. I am leaving tomorrow for a 10-day trip. I'm not sure I will have time to merge this to main before I leave. Should I not, please feel free to include the fixes directly in this PR.

I won't run the workflows until the fixes are in this PR.

I could not thoroughly test building the ktx tool with the above changes because of all the fmt related issues whose fixes are in this PR.

As for the C++23 issue with cppspmd_sse.h, please report it to https://github.com/BinomialLLC/basis_universal/.

MarkCallow · 2026-01-04T08:52:36Z

I'm not sure I will have time to merge this to main before I leave.

The fixes are merged. See PR #1104.

… compile-time checking

ltjax · 2026-01-04T14:47:16Z

Great! It works for C++17 and C++20 for me now, C++23 if I fix or disable (via BASISU_SSE=OFF) the SSE stuff

ltjax · 2026-01-04T14:52:06Z

Interestingly, it seems like you already reported the build problem in basisu a while ago: BinomialLLC/basis_universal#366

MarkCallow reviewed Jan 1, 2026

View reviewed changes

ltjax force-pushed the cpp23 branch from 896bf14 to 82ec8dc Compare January 3, 2026 14:36

ltjax added 3 commits January 4, 2026 15:40

Fix build problems w/ fmt and C++23

043d715

Some more fmt with C++23 compatibility fixes

f51f75a

Replace recursive fmt::format call with nested placeholders to enable…

b6aa7f0

… compile-time checking

ltjax force-pushed the cpp23 branch from e3579ca to b6aa7f0 Compare January 4, 2026 14:42

Fix build problems w/ fmt and C++23 #1102

Are you sure you want to change the base?

Fix build problems w/ fmt and C++23 #1102

Uh oh!

Conversation

ltjax commented Dec 30, 2025

Uh oh!

CLAassistant commented Dec 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MarkCallow left a comment

Choose a reason for hiding this comment

Uh oh!

MarkCallow Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

ltjax Jan 2, 2026 • 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!

MarkCallow Jan 3, 2026

Choose a reason for hiding this comment

Uh oh!

ltjax Jan 3, 2026

Choose a reason for hiding this comment

Uh oh!

MarkCallow commented Jan 2, 2026

Uh oh!

MarkCallow commented Jan 3, 2026

Uh oh!

ltjax commented Jan 3, 2026

Uh oh!

MarkCallow commented Jan 4, 2026

Uh oh!

MarkCallow commented Jan 4, 2026

Uh oh!

ltjax commented Jan 4, 2026

Uh oh!

ltjax commented Jan 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

CLAassistant commented Dec 30, 2025 •

edited

Loading

ltjax Jan 2, 2026 •

edited

Loading