Re: [lmi] Copying compiler run-time files in the makefiles

[Greg Chicares]

On 7/21/22 21:10, Vadim Zeitlin wrote:
> On Thu, 21 Jul 2022 15:02:27 +0000 Greg Chicares <gchicares@sbcglobal.net> 
> wrote:

[...This:

       +@for z in $(compiler_runtime_files); do \
           $(INSTALL) -c -m 0775 -c $$z $(localbindir) ; \
         done;

looks like a goofy way of testing a condition like this:

  [ -z "$(compiler_runtime_files)" ]

and I had thought it was a workaround for a trailing spaced in that
'make' variable; but historically it was a loop that installed these
files to multiple directories, and commit aebe56bb16eae removed the
multiple directories, but retained the goofiness...]

> GC> Maybe
> GC>       [ -z "$(strip $(compiler_runtime_files))" ]
> GC> would work (although I'm not sure I can do that in a recipe),
> 
>  Make functions can be applied anywhere and this looks like a perfectly
> reasonable solution to me.
> 
> GC> but then I'd want to explain why the $(strip ...) function is used.
> 
>  FWIW, I think using strip for the -z argument is rather self-explanatory.

Now that you put it that way, I find that I agree.
 - It looks natural enough.
 - If someone removes it and catastrophe ensues, well...
     http://www.catb.org/jargon/html/D/Don-t-do-that-then-.html
 - Also see the "is natural anyway" comment in the "tests" section below.
And 'install' is designed to work with multiple files, so splitting
them is unnatural. Thus, I'll commit this:

--8<----8<----8<----8<----8<----8<----8<----8<----8<----8<--
diff --git a/GNUmakefile b/GNUmakefile
index 6e99ed135..a27baa18b 100644
--- a/GNUmakefile
+++ b/GNUmakefile
@@ -236,9 +236,8 @@ $(build_dir): $(gpl_files)
        +@[ -d $(localbindir)     ] || $(MKDIR) --parents $(localbindir)
        +@[ -d $(locallibdir)     ] || $(MKDIR) --parents $(locallibdir)
        +@[ -d $(localincludedir) ] || $(MKDIR) --parents $(localincludedir)
-       +@for z in $(compiler_runtime_files); do \
-           $(INSTALL) -c -m 0775 -c $$z $(localbindir) ; \
-         done;
+       +@[ -z "$(strip $(compiler_runtime_files))" ] \
+         || $(INSTALL) -c -m 0775 -c $(compiler_runtime_files) $(localbindir)
        +@$(MAKETARGET)
 
 % :: $(build_dir) ; @:
--8<----8<----8<----8<----8<----8<----8<----8<----8<----8<--

Tests, run with the '@' removed from the beginning of the
replacement line, in an environment where $LMI_TRIPLET
defaults to msw:

/opt/lmi/src/lmi[0]$rm 
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/{libgcc_s_seh-1.dll,libstdc++-6.dll}  
 
/opt/lmi/src/lmi[0]$ls -l 
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/{libgcc_s_seh-1.dll,libstdc++-6.dll}
ls: cannot access 
'/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/libgcc_s_seh-1.dll': No such file or 
directory
ls: cannot access '/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/libstdc++-6.dll': 
No such file or directory

/opt/lmi/src/lmi[2]$make show_flags 2>&1 |grep 'libstdc++'
[ -z "/usr/lib/gcc/x86_64-w64-mingw32/10-win32//libgcc_s_seh-1.dll 
/usr/lib/gcc/x86_64-w64-mingw32/10-win32//libstdc++-6.dll" ] \
  ||  install -c -m 0775 -c 
/usr/lib/gcc/x86_64-w64-mingw32/10-win32//libgcc_s_seh-1.dll 
/usr/lib/gcc/x86_64-w64-mingw32/10-win32//libstdc++-6.dll  
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin

/opt/lmi/src/lmi[0]$ls -l 
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/{libgcc_s_seh-1.dll,libstdc++-6.dll}
-rwxrwxr-x 1 greg lmi   628507 Jul 21 21:55 
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/libgcc_s_seh-1.dll
-rwxrwxr-x 1 greg lmi 26122435 Jul 21 21:55 
/opt/lmi/local/gcc_x86_64-w64-mingw32/bin/libstdc++-6.dll

Same, with a non-msw $LMI_TRIPLET:

/opt/lmi/src/lmi[0]$make show_flags 2>&1 | head --lines=2
[ -z "" ] \
  || install -c -m 0775 -c  /opt/lmi/local/clang_x86_64-pc-linux-gnu/bin

...which demonstrates that $(strip ...) is natural anyway, as it
avoids 'install "" /path/to/wherever', which is invalid no matter
how many blank spaces are between the double quotes.

Commit f760e3df2823, similar but without $(strip ...), mentions that
|    running `make -n fardel` through shellcheck elicits:
|      SC2157: Argument to -z is always false due to literal strings.
so let's try that, just to make assurance doubly sure:

/opt/lmi/src/lmi[1]$echo '#!/bin/sh' > eraseme 
/opt/lmi/src/lmi[0]$make -n show_flags >> eraseme
/opt/lmi/src/lmi[0]$shellcheck eraseme 2>&1 |less -S

Okay, shellcheck is satisfied, too.