fl_scandir && scandir_posix.c

[markjolesen]

if USE_X11 is defined but HAVE_SCANDIR is undefined, you will get a linker error because fl_scandiir is not implemented

[Albrecht-S]

Please tell us your build type (autoconf/make or CMake) and system platform. Is this still on FreeDOS?

Also, please tell us how it comes that USE_X11 is defined on your system, i.e. what environment or build option makes this defined?

[markjolesen]

Yes, it is FreeDOS specific. Consequently, I stick my own stuff in the code so it does not affect me. I am describing a specific use case that appears to be valid. I believe what needs to happen is dummy needs to be replaced with an actual fl_scacndir function that simply returns ::scandir

…

On Tue, Jan 26, 2021 at 12:30 PM Albrecht Schlosser < ***@***.***> wrote: Please tell us your build type (autoconf/make or CMake) and system platform. Is this still on FreeDOS? Also, please tell us how it comes that USE_X11 is defined on your system, i.e. what environment or build option makes this defined? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#183 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA6DEU56DJ7NQCNTFWEYE63S34C4HANCNFSM4WSV6BZA> .

[Albrecht-S]

@markjolesen I'm not sure I understand your statements. I analyzed the code and came to this conclusion (please correct me if I misunderstood something or my conclusion is wrong):

Our only "use case" is to build the FLTK library and fl_scandir() is not part of the public API. You wrote:

if USE_X11 is defined but HAVE_SCANDIR is undefined, you will get a linker error because fl_scandiir is not implemented

That's obviously wrong (maybe a typo?) because the condition is:
#if defined(USE_X11) && !defined(HAVE_SCANDIR)

... but anyway: this is the only condition when fl_scandir() needs to be defined. It is called exactly once in Fl_X11_System_Driver.cxx, line 485:

Fl_X11_System_Driver.cxx-483-#ifndef HAVE_SCANDIR
Fl_X11_System_Driver.cxx-484-  // This version is when we define our own scandir. Note it updates errmsg on errors.
Fl_X11_System_Driver.cxx:485:  int n = fl_scandir(dirloc, list, 0, sort, errmsg, errmsg_sz);
Fl_X11_System_Driver.cxx-486-#elif defined(HAVE_SCANDIR_POSIX)
Fl_X11_System_Driver.cxx-487-  // POSIX (2008) defines the comparison function like this:
Fl_X11_System_Driver.cxx-488-  int n = scandir(dirloc, list, 0, (int(*)(const dirent **, const dirent **))sort);
Fl_X11_System_Driver.cxx-489-#elif defined(__osf__)
. . .

The (otherwise) "empty" #else case at the end of the file has this dummy typedef only to silence a compiler warning. So I am pretty sure there's nothing wrong with the code as-is. Note also that this file is only compiled if USE_X11 is defined (this condition could be removed, it's a leftover from previous versions).

[markjolesen]

I agree that there is something odd about the USE_X11 marco.

…

On Sun, Feb 14, 2021 at 7:13 PM Albrecht Schlosser ***@***.***> wrote: @markjolesen <https://github.com/markjolesen> I'm not sure I understand your statements. I analyzed the code and came to this conclusion (please correct me if I misunderstood something or my conclusion is wrong): Our only "use case" is to build the FLTK library and fl_scandir() is not part of the public API. You wrote: if USE_X11 is defined but HAVE_SCANDIR is undefined, you will get a linker error because fl_scandiir is not implemented That's obviously wrong (maybe a typo?) because the condition is: #if defined(USE_X11) && !defined(HAVE_SCANDIR) ... but anyway: this is the *only* condition when fl_scandir() needs to be defined. It is called *exactly once* in Fl_X11_System_Driver.cxx, line 485: Fl_X11_System_Driver.cxx-483-#ifndef HAVE_SCANDIR Fl_X11_System_Driver.cxx-484- // This version is when we define our own scandir. Note it updates errmsg on errors. Fl_X11_System_Driver.cxx:485: int n = fl_scandir(dirloc, list, 0, sort, errmsg, errmsg_sz); Fl_X11_System_Driver.cxx-486-#elif defined(HAVE_SCANDIR_POSIX) Fl_X11_System_Driver.cxx-487- // POSIX (2008) defines the comparison function like this: Fl_X11_System_Driver.cxx-488- int n = scandir(dirloc, list, 0, (int(*)(const dirent **, const dirent **))sort); Fl_X11_System_Driver.cxx-489-#elif defined(__osf__) . . . The (otherwise) "empty" #else case at the end of the file has this dummy typedef only to silence a compiler warning. So I am pretty sure there's nothing wrong with the code as-is. Note also that this file is only compiled if USE_X11 is defined (this condition could be removed, it's a leftover from previous versions). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#183 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA6DEUZZDZPBAODOPVURDXTS7BYMNANCNFSM4WSV6BZA> .

[Albrecht-S]

OK, I'm leaving everything as-is for now. We will have a check of the new source code WRT such defines and #ifdef's before we release 1.4.0.

I'm closing this with label "wontfix" because there's nothing to "fix", thanks for your report anyway.