bug#28241: man page in r-minimal contains timestamp

[Ricardo Wurmus]

Ricardo Wurmus  writes:

> Ricardo Wurmus  writes:
>
>> Is SOURCE_DATE_EPOCH not set at the time help2man runs?  “r-minimal”
>> uses the gnu-build-system, which has a build phase set-SOURCE-DATE-EPOCH
>> that sets SOURCE_DATE_EPOCH to 1.
>>
>> Where is this lost?
>
> Simple: R simply doesn’t use our help2man.  It comes with
> tools/help2man.pl, which does not respect SOURCE_DATE_EPOCH.
>
> Overwriting the HELP2MAN variable causes the tests to fail, so I need to
> investigate how these versions differ.

This is now fixed with commit 174fbd5f3b2f286cf06e58787853db6b8f0f5b35.

--
Ricardo

GPG: BCA6 89B6 3655 3801 C3C6  2150 197A 5888 235F ACAC
https://elephly.net

bug#28241: man page in r-minimal contains timestamp

[Ricardo Wurmus]

Ricardo Wurmus  writes:

> Is SOURCE_DATE_EPOCH not set at the time help2man runs?  “r-minimal”
> uses the gnu-build-system, which has a build phase set-SOURCE-DATE-EPOCH
> that sets SOURCE_DATE_EPOCH to 1.
>
> Where is this lost?

Simple: R simply doesn’t use our help2man.  It comes with
tools/help2man.pl, which does not respect SOURCE_DATE_EPOCH.

Overwriting the HELP2MAN variable causes the tests to fail, so I need to
investigate how these versions differ.

--
Ricardo

GPG: BCA6 89B6 3655 3801 C3C6  2150 197A 5888 235F ACAC
https://elephly.net

bug#28241: man page in r-minimal contains timestamp

[Ricardo Wurmus]

Ricardo Wurmus  writes:

> The man page is created with a rule in “doc/Makefile.in”:
>
>  @$(HELP2MAN) --include=$(srcdir)/R.aux --no-info --output=R.1 \
>--name="a language for data analysis and graphics" \
>$(top_builddir)/bin/R
>
> “doc/R.aux” does not contain a timestamp, nor does the help output of R
> itself.  Is help2man to blame here?

help2man supports SOURCE_DATE_EPOCH and it does the right thing when it
is set.

--8<---cut here---start->8---
$ guix environment --ad-hoc r-minimal help2man
$ help2man --include=R-3.4.0/doc/R.aux --no-info --output foo.1 R
$ grep August foo.1
.TH R "1" "August 2017" "R version 3.4.0 (2017-04-21) -- "You Stupid Darkness"" 
"User Commands"

$ SOURCE_DATE_EPOCH=0 help2man --include=R-3.4.0/doc/R.aux --no-info --output 
foo.1 R
$ grep August foo.1
$ grep January foo.1
.TH R "1" "January 1970" "R version 3.4.0 (2017-04-21) -- "You Stupid 
Darkness"" "User Commands"
--8<---cut here---end--->8---

Is SOURCE_DATE_EPOCH not set at the time help2man runs?  “r-minimal”
uses the gnu-build-system, which has a build phase set-SOURCE-DATE-EPOCH
that sets SOURCE_DATE_EPOCH to 1.

Where is this lost?

-- 
Ricardo

GPG: BCA6 89B6 3655 3801 C3C6  2150 197A 5888 235F ACAC
https://elephly.net

bug#28241: man page in r-minimal contains timestamp

[Ricardo Wurmus]

The compressed man page at “share/man/man1/R.1.gz” provided by r-minimal
contains a time-dependent string.

Comparing the uncompressed man pages built at different times on
different machines shows me this:

--8<---cut here---start->8---
diff /tmp/berlin/share/man/man1/R.1  /tmp/hydra/share/man/man1/R.1
2c2
< .TH R "1" "July 2017" "R 3.4.0" FSF
---
> .TH R "1" "May 2017" "R 3.4.0" FSF
--8<---cut here---end--->8---

The man page is created with a rule in “doc/Makefile.in”:

 @$(HELP2MAN) --include=$(srcdir)/R.aux --no-info --output=R.1 \
   --name="a language for data analysis and graphics" \
   $(top_builddir)/bin/R

“doc/R.aux” does not contain a timestamp, nor does the help output of R
itself.  Is help2man to blame here?

--
Ricardo

GPG: BCA6 89B6 3655 3801 C3C6  2150 197A 5888 235F ACAC
https://elephly.net