In D30769#691961, @adamw wrote:

Maintainer resets do not require portmgr approval, but I’m happy to give it anyway. When you commit, please reword the approved by line so it doesn’t look like portmgr timed out.

• Remove English only example

It seems all documents that were split has a single/full html in /book/, this is new for me. I noticed after seeing this patch.

In D30718#691109, @blackend wrote:

In D30718#690992, @dbaio wrote:

This is nice to test/build a single document, only issue is that it won't use our template/css.
Just to know, these html's will be used where?

Still rough around the edges, I'm going to add the missing css part.
The aim of this addition is to be able to build one localized version of the docs without the need to build every languages. This is useful for misc/freebsd-doc-* ports.

In D30732#691069, @ceri wrote:

In D30732#690947, @dbaio wrote:

Out of curiosity, is this for help debugging Makefile changes/improvements?

The DOC_LANG restores previous behaviour documented in make.conf (or share/examples/make.conf, I can’t check at the moment) and white space separated variables are arguably easier to work with in make.

The -o option is a start at reintroducing proper dependency tracking, so that the python targets can stop being .PHONY and incremental builds can come back.

Hi...Where can I find the latest diff?

This is nice to test/build a single document, only issue is that it won't use our template/css.
Just to know, these html's will be used where?

This kind of option remembers something with verbose and dry run options.

In D30656#689126, @ceri wrote:

In D30656#689103, @dbaio wrote:

In D30656#689079, @ceri wrote:

I don't understand why BIND and HOSTNAME aren't the same variable?

I imagine people want to bind it to a specific ip/interface and access through a hostname which is different than the ip address.

-b, --baseURL=""    hostname (and path) to the root, e.g.
http://spf13.com/


--bind="127.0.0.1"  interface to which the server will bind

adding @mat, could you please review this as well?

Sorry, to clear, I understand why hugo has them, but I don't really understand why we need them in a development-class Makefile. If we do, then they should be documented so that a human can understand them in the Makefile.

In D30656#689079, @ceri wrote:

I don't understand why BIND and HOSTNAME aren't the same variable?

• Add --minify to HUGO_ARGS variable to make it optional

In D30591#688010, @freebsd_ny-central.org wrote:

In D30591#687963, @dbaio wrote:

There are a lot of patches that changes NSCDE_ROOT to our PREFIX (some do more than that)... What I want to ask is, are there a way to use less patches? For instance, just replacing the variable value?
It will save time on future updates.

And watch out using post-patch target and files/* patches together.

The original install script for nscde puts everything under /opt/nscde; the whole package is set up to be crammed into one directory. Unfortunately, that is not very canonical to how the directory structure under /usr/local is supposed to work.

I had hoped to do it the easy way originally and put everything under /usr/local/nscde or /usr/local/libexec/nscde, but then the documentation would have ended up under something like /usr/local/nscde/share/doc.

Symlinking everything out from that structure kind of felt like cheating, so I distributed things into the locations that matched their use and function (i.e. %DATADIR%, %DOCSDIR%, %ETCDIR% and so on).

Originally, there were even more patches. Fortunately, eduardo@ already pointed out that I respect $PREFIX, which led to some simplification but also meant we still have patch files. Actually, it's a combination:

1. first, we strip those NSCE_ROOT combinations and replace them with $PREFIX and sensible paths
2. then, from the post-patch phase in Makefile we replace "$PREFIX" strings with actual $PREFIX value

If you have any suggestions on how to improve/simplify this or if we should instead go with the approach of putting everyting into /usr/local/nscde (or similar), I'm certainly willing to give it a try.

There are a lot of patches that changes NSCDE_ROOT to our PREFIX (some do more than that)... What I want to ask is, are there a way to use less patches? For instance, just replacing the variable value?
It will save time on future updates.

I've updated docs tree and applied this patch and I'm getting this error when not setting a language:

In D30331#686041, @ceri wrote:

In D30331#683313, @blackend wrote:

In D30331#683208, @ceri wrote:

One other note: the pgpkeys article takes about two hours to build on my (admittedly single-CPU 2GB RAM) machine - is that common and do we need to work around that somehow?

Same here now since we do it right with includes. I just launched a build on my machine (10 years old quad-core) 12 minutes ago and it's still running...
Indeed, we need a solution/workaround.

I've ruled out an include storm by pushing everything into a single file and it still takes over 90 minutes, so will rework the Makefile if I can to avoid as many .PHONY targets as there currently are. No need to hold up this review for it, I think.

Testing here and I've seen this detail on articles/contributors.

Probably form-search-mail needs a similar code as form-search-site.

Just one more thing, about Context not available.
It's nice to generate the diff with full context for reviews.
https://wiki.freebsd.org/Phabricator#Create_a_Revision_via_Web_Interface