Modernize datadir

[zdenop]

• use std::filesystem::path instead of std::string for datadir
• add warning if datadir is not directory or does not exists

[egorpugin]

Probably all is_directory() checks are unnecessary. User will just get any other random error if he points to a file instead of to a dir.

[zdenop]

@stweil : Thank you for review. Unittests needs be fixed yet.

[egorpugin]

Old tprintfs should be removed slowly in favor of std::format.

[stweil]

Old tprintfs should be removed slowly in favor of std::format.

Isn't the tesserr stream even better?

[egorpugin]

Streams are better than tprintf.
But when using them a lot soon you will notice that breaking strings for any insertion is quiet disturbing.

"we add two numbers: " << num1 << " + " << num2 << " = " << num1+num2;

compared to

std::format("we add two numbers: {} + {} = {}", num1, num2, num1+num2);

Having even more insertion points the original string can be hardly readable compared to a string with {} placeholders.

[egorpugin]

Isn't the tesserr stream even better?

I mean with std::format() the usage will be something like tesserr << std::format("format string", args ... );.

[egorpugin]

I've slightly updated tprintf() to C++ style.

8cb0418

[stweil]

I've slightly updated tprintf() to C++ style.

Now I get numerous compiler warnings:

../../../src/ccutil/tprintf.h:37:33: warning: format string is not a string literal (potentially insecure) [-Wformat-security]

And the code size is larger, too.

[egorpugin]

1. Warning should be silenced. Updated code is completely fine and valid. We should remove -Wformat-security flag if set or set -Wno-format-security.
2. What is code size change? In general this should be ignored. But I'm curious what the difference is. Probably 5 bytes for each get_debugfp() and some pre- and post call setups.

[egorpugin]

Can you try changing it to

template <typename ... Types>
auto tprintf(const char *format, Types && ... args) {
  return fprintf(get_debugfp(), format, std::forward<Types>(args)...);
}

?

[stweil]

Can you try changing it to [...]

The suggested change still produces the compiler warnings, but I think it improves the code because like that calling tprintf() without any argument would be caught as an error.

[egorpugin]

It will be an error too if we call tprintf() even without const char *format.

1>tesseract\src\ccutil\tprintf.h(37,10): error C2660: 'fprintf': function does not take 1 arguments
1>(compiling source file '../../../../../src/ccutil/tprintf.cpp')
1>    C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\ucrt\stdio.h(830,37):
1>    see declaration of 'fprintf'
1>    tesseract\src\ccutil\tprintf.h(37,10):
1>    while trying to match the argument list '(FILE *)'
1>    tesseract\src\ccutil\tprintf.h(37,10):
1>    the template instantiation context (the oldest one first) is
1>        tesseract\src\ccutil\tprintf.cpp(74,3):
1>        see reference to function template instantiation 'auto tesseract::tprintf<>(void)' being compiled

[egorpugin]

Added format in d95e9f7

I'll check ci for other warnings.

[egorpugin]

I don't see warnings on gcc-14 (even without format).
Only with clang.

Possible solution is

auto tprintf(const char *format, Types && ... args) {
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wformat-security"
  return fprintf(get_debugfp(), format, std::forward<Types>(args)...);
#pragma clang diagnostic pop
}```

[zdenop]

Seems like unittests GA are failing on the main (too), so it is not related to this PR:

Did I understand it correctly?