Re: [PATCH] configure: Add flags for MinGW32 standalone build

[Joshua Watt]

On Mon, Jan 11, 2021 at 1:44 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 11/01/21 08:29, Stefan Weil wrote:
> > Am 11.01.21 um 08:04 schrieb Thomas Huth:
> >
> >> On 08/01/2021 19.30, Joshua Watt wrote:
> >>>
> >>> On 1/8/21 1:25 AM, Thomas Huth wrote:
> >>>> On 07/01/2021 22.38, Joshua Watt wrote:
> >>>>> There are two cases that need to be accounted for when compiling QEMU
> >>>>> for MinGW32:
> >>>>>   1) A standalone distribution, where QEMU is self contained and
> >>>>>      extracted by the user, such as a user would download from the
> >>>>> QEMU
> >>>>>      website. In this case, all of the QEMU files should be rooted in
> >>>>>      $prefix to ensure they can be easily packaged together for
> >>>>>      distribution
> >>>>>   2) QEMU integrated into a distribution image/sysroot/SDK and
> >>>>>      distributed with other programs. In this case, the provided
> >>>>>      arguments for bindir/datadir/etc. should be respected as they
> >>>>> for a
> >>>>>      Linux build.
> >>>>>
> >>>>> Add a configure time flags --enable-standalone-mingw and
> >>>>> --disable-standalone-mingw that allows the user to control this
> >>>>> behavior. The flag defaults to "enabled" if unspecified to retain the
> >>>>> existing build behavior
> >>>>>
> >>>>> Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
> >>>>> ---
> >>>>>   configure | 8 +++++++-
> >>>>>   1 file changed, 7 insertions(+), 1 deletion(-)
> >>>>>
> >>>>> diff --git a/configure b/configure
> >>>>> index 5860bdb77b..5c83edb502 100755
> >>>>> --- a/configure
> >>>>> +++ b/configure
> >>>>> @@ -358,6 +358,7 @@ strip_opt="yes"
> >>>>>   tcg_interpreter="no"
> >>>>>   bigendian="no"
> >>>>>   mingw32="no"
> >>>>> +mingw32_standalone="yes"
> >>>>>   gcov="no"
> >>>>>   EXESUF="$default_feature"
> >>>>>   HOST_DSOSUF=".so"
> >>>>> @@ -1558,6 +1559,10 @@ for opt do
> >>>>>     ;;
> >>>>>     --disable-fuse-lseek) fuse_lseek="disabled"
> >>>>>     ;;
> >>>>> +  --enable-standalone-mingw) mingw32_standalone="yes"
> >>>>> +  ;;
> >>>>> +  --disable-standalone-mingw) mingw32_standalone="no"
> >>>>> +  ;;
> >>>>>     *)
> >>>>>         echo "ERROR: unknown option $opt"
> >>>>>         echo "Try '$0 --help' for more information"
> >>>>> @@ -1570,7 +1575,7 @@ libdir="${libdir:-$prefix/lib}"
> >>>>>   libexecdir="${libexecdir:-$prefix/libexec}"
> >>>>>   includedir="${includedir:-$prefix/include}"
> >>>>>   -if test "$mingw32" = "yes" ; then
> >>>>> +if test "$mingw32" = "yes" && test "$mingw32_standalone" = "yes";
> >>>>> then
> >>>>>       mandir="$prefix"
> >>>>>       datadir="$prefix"
> >>>>>       docdir="$prefix"
> >>>>> @@ -1897,6 +1902,7 @@ disabled with --disable-FEATURE, default is
> >>>>> enabled if available
> >>>>>     libdaxctl       libdaxctl support
> >>>>>     fuse            FUSE block device export
> >>>>>     fuse-lseek      SEEK_HOLE/SEEK_DATA support for FUSE exports
> >>>>> +  standalone-mingw  Build for standalone distribution on MinGW
> >>>>>     NOTE: The object files are built at the place where configure
> >>>>> is launched
> >>>>>   EOF
> >>>>
> >>>> I think this should maybe be done independently from MinGW, so that
> >>>> it could be used on other systems, too. Thus maybe rather name the
> >>>> switch "--enable-standalone-distribution" or
> >>>> "--enable-standalone-installation" or something like this? On MinGW,
> >>>> the value of the switch could then default to "yes" while on other
> >>>> systems it would be "no" by default.
> >>>
> >>> We could, but I'm curious how useful that is? Does that make the
> >>> option just a shorthand for "--mandir=$prefix --bindir=$prefix
> >>> --datadir=$prefix etc..." for all builds?
> >>
> >> Yes, that would basically be a shorthand for that. Could be useful for
> >> people who want to create standalone binaries on Linux etc., too.
> >>
> >>  Thomas
> >
> >
> > Aren't nearly all files already rooted in $prefix? The only exception I
> > know is /etc/qemu.
> >
> > Rooting in $prefix still allows hierarchical subdirectories. I'd prefer
> > them for MinGW, too.
>
> I agree, it was an issue before 5.2 but now we have relocatable
> installations.  So it would be better to remove all the special casing
> of mingw, except that (for backwards compatibility) on mingw bindir
> defaults to $prefix instead of $prefix/bin.  Then Joshua's usecase is
> covered simply by --bindir=/mingw/bin.

Right. Removing the special case for MinGW would solve my problem and
be fine with me; the thing I don't know is if we need to preserve the
backward compatibility of forcing a specific "flat" layout in $prefix
when building for MinGW. There are basically 2 options I can see:
 1) Add a flag to default MinGW to the "flat" layout, but allow it to
be overridden so that it can follow a normal layout (e.g. --bindir,
etc. are respected). This what my patch adds
 2) Remove the special casing for MinGW entirely. A "flat" layout
would still be possible by doing e.g. "--bindir=$prefix
--datadir=$prefix ..." manually but it would no longer be the default.

I'm fine with either option, but I'm not familiar enough with the
project to be able to know which is better.

>
> Paolo
>