Compiling with GCC 15 gives error too many arguments to function verilogaparse

[papoteur-mga]

gcc -DHAVE_CONFIG_H -I. -I..   -DADMS_INCLUDEDIR=\"/usr/include/adms\"   -O2 -g -pipe -Wformat -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -m64 -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection=full  -Wall -c -o admsXml-admsXml.o `test -f 'admsXml.c' || echo './'`admsXml.c
admsXml.c: In function 'parseva':
admsXml.c:645:8: error: too many arguments to function 'verilogaparse'; expected 0, have 1
  645 |     if(verilogaparse(root()))
      |        ^~~~~~~~~~~~~ ~~~~~~
In file included from admsXml.c:581:
admsVeriloga.h:107:21: note: declared here
  107 | win32_interface int verilogaparse ();
      |                     ^~~~~~~~~~~~~

[kvaadit]

I think this is because of the issue raised here. Basically, when a function signature specifies no arguments (eg, int foo();), GCC-14 by default enforces the gnu17 standard, which allows calling this function with any number of arguments, whereas GCC-15 by default enforces the gnu23 standard, which only allows calling this function with zero arguments.

I was able to get around this by adding -std=gnu17 to my gcc invocations. Or, if you're using CMake, add -DCMAKE_C_FLAGS="-std=gnu17" as a command-line option to your cmake invocation.

But of course, this is only a workaround. The "correct" way to fix this probably involves getting the verilogaparse function signature exactly right in the admsVeriloga.h file.

[felix-salfelder]

adding -std=gnu17 to my gcc invocations.

This should be hardwired in the build system. But it's not. The following seems to work though.

$ ./configure
$ make CFLAGS=-std=gnu17 V=1

I think gnu99 should be sufficient here..

But of course, this is only a workaround.

Not quite. ADMS is not written in C23 and the world has moved on without ADMS.

What is needed is a patch that sets the flag correctly, or better just updates the build instructions.
To go any further, a maintainer is required.

[kvaadit]

I think the right way to address the issue goes beyond a slight tweak to the build system.

Will the code compile with just this tweak? Sure.

But is it good practice to have function prototypes/declarations not match their definitions/invocations? In cases like this, most likely no.

So, it's like GCC-15's default settings have usefully highlighted a slight mistake in the code, which GCC-14 and earlier just used to let slide. That's why I maintain that a "correct" fix should touch the code, not just the build system.

But that's just my opinion. I'm not a maintainer.

[pgajdos]

This fixes the build with gcc15 for me:

Index: ADMS-release-2.3.7/admsXml/admsXml.c
===================================================================
--- ADMS-release-2.3.7.orig/admsXml/admsXml.c
+++ ADMS-release-2.3.7/admsXml/admsXml.c
@@ -642,7 +642,7 @@ static void parseva (const int argc,cons
     if(root()->_dbg_vla==admse_yes)
       adms_veriloga_setint_yydebug(1);
     adms_veriloga_setfile_input(ifh);
-    if(verilogaparse(root()))
+    if(verilogaparse())
       adms_message_fatal(("parse Verilog-AMS preprocessed file '%s' failed\n",mytmpverilogamsfile))
     else
       adms_message_verbose(("No error found during parsing\n"))