Repository navigation
qualify isnan() with std namespace - #565
Merged
Merged
Conversation
When a program that depends on implot includes <implot.h>, we have this sequence of includes: <implot.h> -> <implot_internal.h> -> <imgui_internal.h> -> <math.h> Inside <implot_internal.h>'s source code, there's an unqualified call to isnan(). By doing so, it relies on <imgui_internal.h> having isnan() in the global namespace, which <math.h> is guaranteed to do. However, it's possible that implot's user included <cmath> first, which is not guaranteed to put isnan() in the global namespace, and may prevent a later <math.h> from putting isnan() in the global namespace. On NetBSD 10.0 i386, <cmath> doesn't put isnan() in the global namespace and prevents <math.h> from doing so, causing a compile error. Qualify the call to fix it.
guijan
added a commit
to guijan/dolphin
that referenced
this pull request
Apr 22, 2024
The NetBSD port will only work after epezent/implot#565 is merged and Dolphin updates to a version with the fix. In the meantime, you can apply the fix yourself to a Dolphin checkout.
guijan
added a commit
to guijan/dolphin
that referenced
this pull request
Apr 22, 2024
The NetBSD port will only work after epezent/implot#565 is merged and Dolphin updates to a version with the fix. In the meantime, you can apply the fix yourself to a Dolphin checkout.
Collaborator
|
@guijan is this still an issue with the latest ImGui? I thought |
brenocq
self-requested a review
November 7, 2025 05:34
brenocq
added
type:fix
|
I can confirm that this still happens with master. I just came across this issue while trying to cross-compile from x64 to ARM64 on Linux while using glibc in particular. |
Collaborator
|
Oh... I've moved the issue to high priority :) Are you able to reproduce the issue by modifying the |
|
I've narrowed it down quite a lot, so much so that I can place the reproduction program here: #include <cmath>
#include "implot.h"
#include "implot_internal.h"
int main() { return 0; }Here is what happens when I try to compile this program using glibc and gcc/g++ from x64 to ARM64: Log OutputIt seems including cmath before implot_internal.h causes the issue to show up. This issue does not happen when compiling without a sysroot and cross-compiler, though. |
JoshuaVandaele
added a commit
to JoshuaVandaele/dolphin
that referenced
this pull request
Dec 9, 2025
JoshuaVandaele
added a commit
to JoshuaVandaele/dolphin
that referenced
this pull request
Dec 9, 2025
JoshuaVandaele
added a commit
to JoshuaVandaele/dolphin
that referenced
this pull request
Dec 9, 2025
JoshuaVandaele
added a commit
to JoshuaVandaele/dolphin
that referenced
this pull request
Dec 9, 2025
brenocq
requested changes
Dec 21, 2025
Collaborator
Yep, that looks pretty narrow to me... |
brenocq
added
comment:wontfix
brenocq
added
type:fix
Collaborator
|
I was planning to close this issue since I can't reproduce it, but I've got an idea. Can someone (@guijan @JoshuaVandaele) try to add #ifndef IMPLOT_VERSION
#error Must include implot.h before implot_internal.h
#endif
#ifndef IMGUI_DISABLE
#include <time.h>
+#include <math.h>
#include "imgui_internal.h"
// Support for pre-1.84 versions. ImPool's GetSize() -> GetBufSize()
#if (IMGUI_VERSION_NUM < 18303)
#define GetBufSize GetSize
#endif
|
Sadly this does not fix the issue. Using either |
std::isnan alone still fails in the configuration that broke the original isnan: when the C library's <math.h> is found before the C++ standard library's (a sysroot or system include directory passed with -I), isnan stays a macro and std::isnan expands to the invalid std::__builtin_isnan, and with libc++ std::isnan is not declared at all unless <cmath> is included first. Including <cmath> ourselves is not enough either, since libc++ rejects that include order outright. ImNan and ImNanOrInf now copy the double into an ImU64 and test the IEEE-754 exponent and mantissa, which needs no <cmath> and no isnan macro or function. Unlike isnan, std::isnan or val != val, the bit test also keeps working under -ffast-math and /fp:fast, where the compiler may fold floating-point NaN checks to false and NaN gaps in plots silently disappear. Under -ffinite-math-only an empty asm statement hides the bits from the optimizer, because clang otherwise folds the exponent test to false as well.
Collaborator
|
What was wrong. What the fix does. I added a commit on top of yours that stops asking the compiler "is this NaN?". It reads the 64 bits of the static inline bool ImNan(double val) { return (ImDoubleBits(val) & 0x7FFFFFFFFFFFFFFFull) > 0x7FF0000000000000ull; }This needs no math header, so include order doesn't matter. GCC treats it as plain integer math, so fast-math leaves it alone. Clang would still fold it away, so the commit adds a tiny optimizer barrier that is only active in fast-math builds. How I tested it.
I'll port the same change to ImPlot3D. Let me know if you guys still experience the issue! |
oliverdixon
pushed a commit
to oliverdixon/implot3d
that referenced
this pull request
Nov 2, 2026
ImNan called the unqualified isnan, which fails to build when <cmath> is included before ImPlot3D and the C library's <math.h> is found ahead of the C++ one (a sysroot or system include directory passed with -I): isnan is then neither a macro nor a global function. Switching to std::isnan only moves the failure to other setups, such as libc++ without <cmath>. ImNan and ImNanOrInf now copy the double into an ImU64 and test the IEEE-754 exponent and mantissa, so they need no math header, and unlike isnan or val != val they keep detecting NaN under -ffast-math and /fp:fast, where the compiler may delete floating-point NaN checks. Under -ffinite-math-only an empty asm statement hides the bits from the optimizer, since clang otherwise folds the exponent test to false. Mirrors the same fix in ImPlot (epezent/implot#565).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See the commit message for details. I've tested this change on NetBSD only.
Found while porting dolphin-emu/dolphin to NetBSD: