Re: [PATCH v2 01/30] configure: Allow user to specify sphinx-build binary

[Markus Armbruster]

Peter Maydell <address@hidden> writes:

> On Fri, 14 Feb 2020 at 06:33, Markus Armbruster <address@hidden> wrote:
>>
>> Does not work out of the box on my Fedora 30 build host, where
>> sphinx-build gives me sphinx-build-2.  I have to specify
>> --sphinx-build=/usr/bin/sphinx-build-3 to unbreak it.  Which of course
>> breaks things when I try to build anything before this commit
>>
>> The appended patch makes it work out of the box.  Please consider
>> squashing it in.
>>
>> diff --git a/configure b/configure
>> index 14172909f0..a9d175c400 100755
>> --- a/configure
>> +++ b/configure
>> @@ -584,7 +584,6 @@ query_pkg_config() {
>>  }
>>  pkg_config=query_pkg_config
>>  sdl2_config="${SDL2_CONFIG-${cross_prefix}sdl2-config}"
>> -sphinx_build=sphinx-build
>>
>>  # If the user hasn't specified ARFLAGS, default to 'rv', just as make does.
>>  ARFLAGS="${ARFLAGS-rv}"
>> @@ -903,6 +902,7 @@ fi
>>
>>  : ${make=${MAKE-make}}
>>  : ${install=${INSTALL-install}}
>> +
>>  # We prefer python 3.x. A bare 'python' is traditionally
>>  # python 2.x, but some distros have it as python 3.x, so
>>  # we check that too
>> @@ -915,6 +915,19 @@ do
>>          break
>>      fi
>>  done
>> +
>> +set -x
>
> I guess the set -x / set +x here are accidentally left in
> debug printing?

Mispasted.  I just double-checked these two lines are the only crap I
left in.

>> +sphinx_build=
>> +for binary in sphinx-build-3 sphinx-build
>> +do
>> +    if has "$binary"
>> +    then
>> +        sphinx_build=$(command -v "$binary")
>> +        break
>> +    fi
>> +done
>> +set +x
>> +
>>  : ${smbd=${SMBD-/usr/sbin/smbd}}
>>
>>  # Default objcc to clang if available, otherwise use CC
>> @@ -4803,7 +4816,7 @@ has_sphinx_build() {
>>      # sphinx-build doesn't exist at all or if it is too old.
>>      mkdir -p "$TMPDIR1/sphinx"
>>      touch "$TMPDIR1/sphinx/index.rst"
>> -    $sphinx_build -c "$source_path/docs" -b html "$TMPDIR1/sphinx" 
>> "$TMPDIR1/sphinx/out" >/dev/null 2>&1
>> +    "$sphinx_build" -c "$source_path/docs" -b html "$TMPDIR1/sphinx" 
>> "$TMPDIR1/sphinx/out" >/dev/null 2>&1
>>  }
>
> This change isn't related to trying sphinx-build-3 --
> did you actually need it ?

If the for loop finds nothing, $sphinx_build remains empty.  Quoting the
variable seems cleaner.

Oh, and if the user passes '--sphinx-build=', $sphinx_build becomes
empty.  Precedes my fixup.  Admittedly a rather silly thing to do.

> I did think about quoting when I wrote the patch,
> but looking at existing practice we are all over the
> place on whether we bother to quote variables containing
> program names in configure. I think I ended up following
> the same thing we do for $python, which doesn't quote.

I tend to omit quotes when it's obvious the variable's value can only be
harmless.  An empty value isn't.

> Other than that, I'm happy to squash this in, or for
> you to squash it in if you are otherwise OK taking
> the first chunk of the patchset via your tree now.
> (Do you have a preference for whether you take these
> patches via your tree or I send them in a docs pullreq?)

I can do the pull request.