Re^4: Defining an XS symbol in the Makefile.PL (largely solved)

[jcb]

Are the % characters appearing in the generated Makefile where they should be?

When make prints the command it is issuing, is the % present as it should be?

If it is in the Makefile and missing when make reports the command, file a bug on GNU make, but if GNU make prints the %, my guess is that the Windows shell is eating the single %, a situation that dmake silently works around by doubling % characters (or something else; if the Windows shell is responsible for this, your solution of writing %% may only work due to Yet Another Microsoft Bug).

If you try installing bash and editing the generated Makefile to set SHELL = /path/to/bash.exe, do the strange problems go away? If so, it is almost a Windows-ism tripping you up here.

GNU usually has consistent escaping and quoting rules. I doubt that GNU make is that different between "native" and "Windows" ports. Have you tried a Cygwin build of GNU make?

Unless your module is intended to be Windows-only, I suggest that the most portable solution is to avoid passing format string constants on the command line.

[syphilis]

Are the % characters appearing in the generated Makefile where they should be?

Yes - if I've specified one "%", then only one appears in the generated Makefile - irrespective of the flavour of make.
If I've specified "%%", then "%%" appears in the generated Makefile - irrespective of the flavour of make.

When make prints the command it is issuing, is the % present as it should be?

Yes - according to what's appearing on the screen as the build is proceeding, the actual command that's being run is as expected for the particular value I've given to CCFLAGSEX, irrespective of the flavour of make.

There were problems (with which I did not persevere) in getting that Inline::C script to run in my (MSYS2) bash shell using my Windows perl. The process hangs (due to issues with backslashes in paths, I think):

$ perl try.pl
validate Stage
Starting Build Preprocess Stage
get_maps Stage
Finished Build Preprocess Stage

Starting Build Parse Stage
Finished Build Parse Stage

Starting Build Glue 1 Stage
Finished Build Glue 1 Stage

Starting Build Glue 2 Stage
Finished Build Glue 2 Stage

Starting Build Glue 3 Stage
Finished Build Glue 3 Stage

Starting Build Compile Stage
  Starting "perl Makefile.PL" Stage
Generating a gmake-style Makefile
Writing Makefile for try_pl_d3118
Writing MYMETA.yml and MYMETA.json
  Finished "perl Makefile.PL" Stage

  Starting "make" Stage
Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation.  All rights reserved.

C:\_32\C\_Inline\build\try_pl_d3118> [Process hangs here]

However, I've just found another Windows GNU make (provided by Strawberry Perl):

GNU Make 4.2.1
Built for x86_64-w64-mingw32
Copyright (C) 1988-2016 Free Software Foundation, Inc.

and using that particular build of make makes the issue go away !!!
That's a good enough indicator for me that my problems arose because I was using a buggy version of GNU make.

Thanks for the continuing assistance, guys.
Further conversation/thoughts/observations are certainly welcome, but I'm now marking this as solved.

Cheers,
Rob

[jcb]

Considering that your Inline::C test is hanging just after producing a cmd.exe startup banner and prompt, I would file this whole mess under the category of "Windows": evidently make is starting a shell to run a command but the shell is waiting for input (probably from NUL:) instead of running the intended command.

[syphilis]

To me, it just illustrates that one shouldn't execute a native windows perl.exe inside an MSYS2 bash shell.
That it failed to run to completion in that shell is not worthy of any concern.

Cheers,
Rob

[Anonymous Monk]

jcb: windows cmd.exe doesn't eat % randomly