gpakosz / ppk_assert Goto Github PK

"The library is designed to be lightweight would you decide to keep assertions enabled even in release builds (#define PPK_ASSERT_ENABLED 1)"

Unless I don't know something about English language, you need some joining word before "would". Also, the use of "would" is a bit weird here.
It should be something like
"The library is designed to be lightweight so that you can keep assertions enabled even in release builds (#define PPK_ASSERT_ENABLED 1)"

Currently, one must decide at compile time whether PPK_ASSERT will use a static bool _ignore flag or not (see: PPK_ASSERT_DISABLE_IGNORE_LINE).

However, I have a program that largely should use that bool, but can't in certain places.
The place where it is not allowed is inside a constexpr function.

eg:

constexpr int myFunc(int arg) {
    PPK_ASSERT(arg != 0);
    /* ... */
    return arg+5;
}

This fails to compile for me, with:

ppk_assert.h line 243 col 23: error: ‘_ignore’ declared ‘static’ in ‘constexpr’ function

So, it would be nice if there were some PPK_ASSERT_NOSTATE(...) set of macros that could be used in this case, even if the normal PPK_ASSERT(...) macros do keep state.

In the situation where the assert wants to break into the debugger, there is currently some hardcoded choice of using __debugbreak, raise(SIGTRAP), etc. depending on platform.

This is not bad, but it would be nice to let the user #define their own handling for that situation.

For instance, one could use the debugbreak code, which has some more nuanced platform checks, and can use int3 and other methods.

I got this warning when building with gcc for release (not debug, not with clang):

In file included from /usr/include/string.h:494,
                 from /usr/include/c++/10/cstring:42,
                 from ../engine/third-party/PPK_ASSERT/src/ppk_assert.cpp:13:
In function ‘char* strncpy(char*, const char*, size_t)’,
    inlined from ‘ppk::assert::AssertionException& ppk::assert::AssertionException::operator=(const ppk::assert::AssertionException&)’ at ../engine/third-party/PPK_ASSERT/src/ppk_assert.cpp:386:16:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:106:34: error: ‘char* __builtin_strncpy(char*, const char*, long unsigned int)’ output truncated before terminating nul copying as many bytes from a string as its length [-Werror=stringop-truncatio]
  106 |   return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
      |          ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../engine/third-party/PPK_ASSERT/src/ppk_assert.cpp: In member function ‘ppk::assert::AssertionException& ppk::assert::AssertionException::operator=(const ppk::assert::AssertionException&)’:
../engine/third-party/PPK_ASSERT/src/ppk_assert.cpp:351:27: note: length computed here
  351 |     size_t length = strlen(message);
      |                     ~~~~~~^~~~~~~~~

Note that I enabled warnings as errors, so this prevents compilation for me, but otherwise it would just be a warning.

I found this thread https://stackoverflow.com/questions/65127257/gcc-strncat-output-may-be-truncated-werror-stringop-truncation-what-does suggesting to use memcpy, but since I don't want to modify third-party code and I want to keep all warnings as errors, I just disabled the stringop-truncation entirely (note it's written as "stringop-truncatio" in the message above).

I personally prefer the more semantical strncpy, so I would understand if you didn't want to use memcpy or use some other tricks to remove the warning. Of course, if you have a clean solution, just go ahead. I looked at the code and the lengths seem correct, but as length is computed dynamically, the compiler may not recognize that.

Even if you don't change your code, I put this issue here so people having the same problem know about some solutions.

my guess is - if I want to build ppk_assert as a .dll?

Thanks for the great project!

And a side question - would it be possible to remove as much as possible from the source file if PPK_ASSERT_ENABLED is 0? I don't want to rely on link-time optimizations to remove dead code.

And one more note: Since C++11 local static variables are required to have synchronization added by the compiler for their initialization - so calling the function from multiple threads is safe. This may add A LOT of overhead if asserts are used on hot paths... Maybe this should be mentioned near PPK_ASSERT_DISABLE_IGNORE_LINE in the docs? I will disable this from day 1 in my code...

And one more thing: switching from #include <ppk_assert.h> to #include "ppk_assert.h" would be nice - so the folder doesn't have to be in the include path.

And one more thing : stuff like PPK_ASSERT_ACTION_BREAK from the documentation isn't present in the header - it has been moved to an enum.

Whops - I added quite a few things in this single "issue" :D

Hi there!

I am trying to use your simple static_assert macro found on this site:
http://pempek.net/articles/2013/11/23/boost-static-assert-without-boost/
It is working with gcc 4.3.4

However I tied with 4.8.2 and got the following:

error: expected \ufffd\ufffd\ufffd)\ufffd\ufffd\ufffd before \ufffd\ufffd\ufffd,\ufffd\ufffd\ufffd token
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: expected unqualified-id before numeric constant
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: \ufffd\ufffd\ufffd__static_assertion_at_line_\ufffd\ufffd\ufffd was not declared in this scope
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: invalid use of incomplete type \ufffd\ufffd\ufffdstruct JOIN\ufffd\ufffd\ufffd
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: forward declaration of \ufffd\ufffd\ufffdstruct JOIN\ufffd\ufffd\ufffd
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: template argument 1 is invalid
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: invalid type in declaration before \ufffd\ufffd\ufffd(\ufffd\ufffd\ufffd token
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");
error: typedef \ufffd\ufffd\ufffdJOIN\ufffd\ufffd\ufffd is initialized (use decltype instead)
error: \ufffd\ufffd\ufffd__static_assertion_test_at_line_\ufffd\ufffd\ufffd was not declared in this scope
 STATIC_ASSERT(false, "no_you_cant_have_a_pony");

Do you have some simple static_assert code which is working on 4.8.2 - any idea what could be the problem?
Thanks in advance!
Balazs