Message-ID: <CAEMqcmp8hRuObEg1p5jLY0VqW5-vjm-6RC5t2x_JNcRrNo0y8A@mail.gmail.com>
Date:	Tue, 14 Jul 2015 06:35:42 +0200
From:	Ulf Magnusson <ulfalizer.lkml@...il.com>
To:	Jim Davis <jim.epost@...il.com>
Cc:	linux-kernel <linux-kernel@...r.kernel.org>,
	linux-doc <linux-doc@...r.kernel.org>,
	Michal Marek <mmarek@...e.cz>,
	Jonathan Corbet <corbet@....net>,
	Herbert Xu <herbert@...dor.apana.org.au>, smueller@...onox.de,
	Ulf Magnusson <ulfalizer@...il.com>
Subject: Re: [PATCH] DocBook: Avoid stdout junk with no man pages to compress

On Tue, Jul 14, 2015 at 6:27 AM, Ulf Magnusson <ulfalizer.lkml@...il.com> wrote:
> On Mon, Jul 13, 2015 at 2:46 AM, Ulf Magnusson <ulfalizer.lkml@...il.com> wrote:
>> On Sun, Jul 12, 2015 at 04:36:53PM -0700, Jim Davis wrote:
>>> On Sun, Jul 12, 2015 at 2:59 PM, Ulf Magnusson <ulfalizer.lkml@...il.com> wrote:
>>> > gzip would run as 'gzip -f' when no uncompressed man pages were found,
>>> > making it compress the (empty) stdin to stdout.
>>>
>>> > --- a/Documentation/DocBook/Makefile
>>> > +++ b/Documentation/DocBook/Makefile
>>> > @@ -56,7 +56,7 @@ htmldocs: $(HTML)
>>> >
>>> >  MAN := $(patsubst %.xml, %.9, $(BOOKS))
>>> >  mandocs: $(MAN)
>>> > -       find $(obj)/man -name '*.9' | xargs gzip -f
>>> > +       find $(obj)/man -name '*.9' -exec gzip -f {} \;
>>> >
>>> >  installmandocs: mandocs
>>> >         mkdir -p /usr/local/man/man9/
>>>
>>> That does get rid of the binary burp, but 'xargs gzip -f'  has been in
>>> the Makefile since January, and gzipping '\n' just started recently.
>>> So what's changed?
>>>
>>
>> No idea. I just assumed it had been broken since then, since the version
>> before d56fcf299fb4 (DocBook: Do not exceed argument list limit) looked
>> for *.9 files before running gzip:
>>
>> mandocs: $(MAN)
>>         $(if $(wildcard $(obj)/man/*.9),gzip -f $(obj)/man/*.9)
>>
>>> It looks like, for whatever reason, make installmandocs always ends up
>>> rerunning mandocs -- there's now a 'GEN  Documentation
>>> Docbook//v4l2.xml' printed, and that extra mandocs invocation is where
>>> the problematic second invocation of find is coming from.  I won't
>>> pretend to understand the Makefile flow to guess at why that's
>>> happening, but obviously 'make mandocs; make installmandocs' shouldn't
>>> need to regenerate things already generated.
>>
>> I won't pretend to understand the Makefile flow either. Guess it might
>> be worth looking into v4l2.xml as well then. Could be some directory
>> shenanigans going on judging from the '//'.
>>
>
> I looked into it some more, and I'm now fairly certain that 'mandocs' always
> runs its recipe regardless of the status of the prerequisites. Changing the
> top-level Makefile to do
>
>         $(Q)$(MAKE) --debug=v $(build)=Documentation/DocBook $@
>
> and running 'make mandocs', you see some FORCEs in the output, which -- unless
> my make fu is weak -- causes all the targets above that to always be considered
> out of date:
>
> File 'mandocs' does not exist.
>  Considering target file 'Documentation/DocBook/z8530book.9'.
>    Considering target file 'Documentation/DocBook/z8530book.xml'.
>      ...
>      Considering target file 'FORCE'.
>       File 'FORCE' does not exist.
>       Finished prerequisites of target file 'FORCE'.
>      Must remake target 'FORCE'.
>      Successfully remade target file 'FORCE'.
>      ...
>    Must remake target 'Documentation/DocBook/z8530book.xml'.
>  ...
> Must remake target 'mandocs'.

Had forgotten that 'mandocs' is a phony target too. That alone makes it always
run its recipe when it's a prerequisite of some target that's run.

>
>
> Re. the 'GEN  Documentation Docbook//v4l2.xml', I think the problem is the
> following rule in Documentation/DocBook/media/Makefile:
>
> $(MEDIA_OBJ_DIR)/v4l2.xml: $(OBJIMGFILES)
>         @$($(quiet)gen_xml)
>         @(ln -sf `cd $(MEDIA_SRC_DIR) && /bin/pwd`/v4l/*xml $(MEDIA_OBJ_DIR)/)
>         @(ln -sf `cd $(MEDIA_SRC_DIR) && /bin/pwd`/dvb/*xml $(MEDIA_OBJ_DIR)/)
>
> ./Documentation/DocBook/v4l2.xml is a symlink, so make compares the
> modification time of the symlink *target* (which is never updated) against the
> image (*.png, *.gif, etc.) files in $(OBJIMGFILES). Updating the symlink itself
> won't change that modification time, so that's why it always runs.
>
> Passing --check-symlink-times to make so that it also looks at the modification
> time of the symlink seems to change a bunch of other stuff in the output too.
> Maybe there's other problems lurking here as well.
>
>>>
>>> In any event,
>>>
>>> Tested-by: Jim Davis <jim.epost@...il.com>
>>>
>>> Jim
>>
>> I just noticed the commit message only mentions the alternative
>> solutions and not the implemented solution. Could send a v2 that fixes
>> that, but I'll wait for more comments first.
>>
>> Cheers,
>> Ulf
>
> Cheers,
> Ulf

Cheers,
Ulf
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists