In D27954#624728, @grembo wrote:

In D27954#624718, @mat wrote:

I don't think this is a good idea. Having REINPLACE_ARGS overridable was probably a bad idea, and the ability to change -i is also something that needs to go away.

@mat Well, why don't we commit this to fix what was actually broken in r522484 (this one-line patch is absolutely non-intrusive and doesn't make anything worse) to stop people from running into the problem now and you come up with and execute a plan to retire overriding REINPLACE_ARGS as a separate effort?

I don't think this is a good idea. Having REINPLACE_ARGS overridable was probably a bad idea, and the ability to change -i is also something that needs to go away.

I don't understand why so many ports are changed.

In D27340#611228, @adamw wrote:

The "Any of:" part doesn't feel right to me, because it's restrictive. Can you maybe flip it around? "Applications, utilities, or games primarily[...] or study in general, such as: course-writing or course-delivery applications[...]". Something like that gives context and hints as to what belongs without restricting its ability to grow over time.

• Ignore grep return value.

• fixup! Add set pipefail in most framework scripts.

Can you also convert a few ports to this after you add the keyword?

ok by me.

I am sure there are examples of ports doing that, but listing a few here would be nice.

In D26558#592594, @kevans wrote:

In D26558#592593, @mat wrote:

In D26558#592562, @kevans wrote:

In D26558#592498, @mat wrote:

I disagree with those changes, x11/gnome-pie is not using REINPLACE_ARGS correctly. That variable can only be used to set the -i argument.

Sure. Do we document this somewhere? We should, at or around fixing that particular usage.

REINPLACE_ARGS is purposefully not documented in the porter's handbook.
An extra comment in Mk/bsd.port.mk would not hurt though.

Sure, I can get behind that, maybe:

# REINPLACE_ARGS may only be used to set or override the -i argument.  Any other use is considered
# invalid.

?

In D26558#592562, @kevans wrote:

In D26558#592498, @mat wrote:

I disagree with those changes, x11/gnome-pie is not using REINPLACE_ARGS correctly. That variable can only be used to set the -i argument.

Sure. Do we document this somewhere? We should, at or around fixing that particular usage.

I disagree with those changes, x11/gnome-pie is not using REINPLACE_ARGS correctly. That variable can only be used to set the -i argument.

In D24240#568481, @mandree wrote:

No more complaints received in a month, can I commit?

Assuming this was tested on all our supported releases.

Assuming you have tested it and it works.

I am not sure if it is the correct way to fix this either, but note that the src component has a special treatment, for some reason, and also, if you install FreeBSD without the src component from the installer, the resulting freebsd-update.conf file still has the src component,

Well, the bureaucrat in me thinks "open a PR, wait 14 days, commit". ;-)

In D26221#583669, @jrm wrote:

In D26221#583268, @mat wrote:

Why is it quoted in the first place, and why is it broken now?

It looks like it's quoted because in Mk/Uses/python.mk the package list is generated using PYDISTUTILS_INSTALLARGS:= --record ${_PYTHONPKGLIST}. It's broken because different things under Mk/ do not expect a quote, e.g., ${SED} -e 's|^${PREFIX}/||' ${_PYTHONPKGLIST} ${TMPPLIST} or when doing stage-qa.

Why is it quoted in the first place, and why is it broken now?