lld ignores other arguments when `-V` is passed

[luke-jr]

lld -V <anything> results in the expected behaviour of -V with everything else ignored, exiting with 0.

It appears GNU ld will run -V and then proceed as normal with the rest of the command line. cmake appears to (sometimes?) rely on this behaviour presently (cmake issue 23969).

I'm not sure if lld wants to imitate ld's behaviour or this is a works-as-intended.

[luke-jr]

This breaks gcc -v ("Print (on standard error output) the commands executed to run the stages of compilation. Also print the version number of the compiler driver program and of the preprocessor and the compiler proper.")

[luke-jr]

Looks like a trivial fix:

--- a/lld/ELF/Driver.cpp
+++ b/lld/ELF/Driver.cpp
@@ -531,9 +531,7 @@ void LinkerDriver::linkerMain(ArrayRef<const char *> argsArr) {
 
   // The behavior of -v or --version is a bit strange, but this is
   // needed for compatibility with GNU linkers.
-  if (args.hasArg(OPT_v) && !args.hasArg(OPT_INPUT))
-    return;
-  if (args.hasArg(OPT_version))
+  if ((args.hasArg(OPT_v) || args.hasArg(OPT_version)) && !args.hasArg(OPT_INPUT))
     return;
 
   // Initialize time trace profiler.

[llvmbot]

@llvm/issue-subscribers-lld-elf

[MaskRay]

This breaks gcc -v ("Print (on standard error output) the commands executed to run the stages of compilation. Also print the version number of the compiler driver program and of the preprocessor and the compiler proper.")

Can you be more specific which part of gcc -v breaks? (Please summarize your CMake issue discussion). gcc -v doesn't pass -v to the linker.

I can confirm that without an input, both -v and --version exit without producing an output, but it's unclear what is a required behavior. And there is something weird like:

% ld.bfd -v -v --version --version
GNU ld (GNU Binutils for Debian) 2.38.90.20220713
GNU ld (GNU Binutils for Debian) 2.38.90.20220713
GNU ld (GNU Binutils for Debian) 2.38.90.20220713
Copyright (C) 2022 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License version 3 or (at your option) a later version.
This program has absolutely no warranty.

ld(1) merely says:

       -v
       --version
       -V  Display the version number for ld.  The -V option also lists the supported emulations.

I think the usual expectation of a command line utility is that a style --version should suppress normal output, so the behavior is weird.

I think we can still make the suggested change since it doesn't seem to increase complexity.

[luke-jr]

gcc -v passes -V (uppercase, alias for --version) to lld (at least on ppc64le).

[MaskRay]

gcc -v passes -V (uppercase, alias for --version) to lld (at least on ppc64le).

Thanks. So it seems that for many targets (e.g. many *-freebsd and powerpc*) GCC have such a default LINK_SPEC %{v:-V}, i.e. gcc -v passes -V to ld.
This seems a convincing argument to emulate this behavior, even if it is weird.
We have to treat -V and --version differently since --version needs to stop processing from the GNU ld behavior (if any project relies upon this behavior, it's probably libtool or autotools).

[robertmaynard]

To be clear this isn't a freebsd/powerpc quirk. Commands invocations like g++ source.cpp -Xlinker -V should be valid on all platforms for linking when you also want to get verbose link output ( done since -V gives more info compared to -v with ld).

[MaskRay]

To be clear this isn't a freebsd/powerpc quirk. Commands invocations like g++ source.cpp -Xlinker -V should be valid on all platforms for linking when you also want to get verbose link output ( done since -V gives more info compared to -v with ld).

Whether driver -v should forward -V to as/ld is debatable. https://sourceware.org/pipermail/binutils/2022-September/123031.html "The more flags like this we force pass to the additional tools the more likely we are to have problems when that tool is not from the GNU toolchain."

I agree with that opinion doing this is just unnecessary.