Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-23 Thread Philip Rinn
Hi Dirk,

thanks for your fast upload. Now only 10 out of 268 tested GNU R packages FTBR. 
That's a huge step forward!

And only for three of them it seems to be caused by GNU R - their {PACKAGE}.rdb 
and {PACKAGE}.rdx differ:

r-cran-amelia
https://reproducible.debian.net/dbd/unstable/amd64/r-cran-amelia_1.7.3-1.debbindiff.html

r-cran-rsdmx
https://reproducible.debian.net/dbd/unstable/amd64/r-cran-rsdmx_0.4.5+dfsg-1.debbindiff.html

r-cran-teachingdemos
https://reproducible.debian.net/dbd/unstable/amd64/r-cran-teachingdemos_2.9-1.debbindiff.html

I'll investigate further and report a bug if I find something.

Best,
Philip


signature.asc
Description: Digital signature


Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-19 Thread Philip Rinn
On 18.04.2015 at 15:37, Dirk Eddelbuettel wrote:
 
 On 18 April 2015 at 15:22, Johannes Ranke wrote:
 | Am Samstag, 18. April 2015, 06:27:03 schrieb Dirk Eddelbuettel:
 |  On 18 April 2015 at 11:15, Philip Rinn wrote:
 |  | Hi,
 |  | 
 |  | I think this patch is more what I wanted.
 |  | It now sets the build-timestamp to the time of the last changelog entry 
 if
 |  | the maintainer does not set builttimeStamp explicitly.
 |  | With this change almost all GNU R packages in Debian would automatically
 |  | build reproducible.
 |  
 |  And I presume that is what we want?  I never heard from the 
 reproducibility
 |  group after I made the two patches (cf old bug report) to R itself and
 |  r-cran.mk.

Yes, I think so. I asked on the #debian-reproducible IRC channel yesterday and 
got
a looks good to me back.
This change would make at least 223 [1] GNU R packages build reproducible - 
that's
about 5% of all FTBR packages.

 | Without having further background information I would say that the build-
 | timestamp should be either set to the build time (to carry correct 
 | information) or not at all (to have reproducible builds).
 
 Context: R _always_ puts one hin, hence the discussion in the earlier bug
 report and my (long-accepted) upstream ptahc to be able to override.
 
 What Philip is suggesting here is to use the hook I provided with a suitable
 value -- the timestamp from debian/changelog.  I think I'll do that.

Well, defaulting to builttimeStamp =   would work too, but for me it looks
better to have the timestamp when the package was modified there.

 Dirk, at a workshop
  
 | Johannes
 | 
 |  Too bad you sent me this today. I just made four uploads of R in the last
 |  few days leading up to R 3.2.0.

Yes, I saw that, but I didn't had time earlier this week. Having that set with 
the
start of the new release cycle would have been nice but it's not mandatory I 
think.

Philip

 |  
 |  Dirk
 |  
 |  | Best,
 |  | Philip
 |  | 
 |  | x[DELETED ATTACHMENT usable_builttimeStamp2.patch, text/x-patch]
 |  | x[DELETED ATTACHMENT signature.asc, application/pgp-signature]
 

[1]
https://reproducible.debian.net/issues/unstable/timestamps_in_description_files_generated_by_r-base-dev_issue.html



signature.asc
Description: OpenPGP digital signature


Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-18 Thread Johannes Ranke
Am Samstag, 18. April 2015, 06:27:03 schrieb Dirk Eddelbuettel:
 On 18 April 2015 at 11:15, Philip Rinn wrote:
 | Hi,
 | 
 | I think this patch is more what I wanted.
 | It now sets the build-timestamp to the time of the last changelog entry if
 | the maintainer does not set builttimeStamp explicitly.
 | With this change almost all GNU R packages in Debian would automatically
 | build reproducible.
 
 And I presume that is what we want?  I never heard from the reproducibility
 group after I made the two patches (cf old bug report) to R itself and
 r-cran.mk.

Without having further background information I would say that the build-
timestamp should be either set to the build time (to carry correct 
information) or not at all (to have reproducible builds).

Johannes

 Too bad you sent me this today. I just made four uploads of R in the last
 few days leading up to R 3.2.0.
 
 Dirk
 
 | Best,
 | Philip
 | 
 | x[DELETED ATTACHMENT usable_builttimeStamp2.patch, text/x-patch]
 | x[DELETED ATTACHMENT signature.asc, application/pgp-signature]


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-18 Thread Dirk Eddelbuettel

On 18 April 2015 at 11:15, Philip Rinn wrote:
| Hi,
| 
| I think this patch is more what I wanted.
| It now sets the build-timestamp to the time of the last changelog entry if the
| maintainer does not set builttimeStamp explicitly.
| With this change almost all GNU R packages in Debian would automatically build
| reproducible.

And I presume that is what we want?  I never heard from the reproducibility
group after I made the two patches (cf old bug report) to R itself and 
r-cran.mk.

Too bad you sent me this today. I just made four uploads of R in the last few
days leading up to R 3.2.0.

Dirk
 
| Best,
| Philip
| 
| x[DELETED ATTACHMENT usable_builttimeStamp2.patch, text/x-patch]
| x[DELETED ATTACHMENT signature.asc, application/pgp-signature]

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-18 Thread Dirk Eddelbuettel

On 18 April 2015 at 15:22, Johannes Ranke wrote:
| Am Samstag, 18. April 2015, 06:27:03 schrieb Dirk Eddelbuettel:
|  On 18 April 2015 at 11:15, Philip Rinn wrote:
|  | Hi,
|  | 
|  | I think this patch is more what I wanted.
|  | It now sets the build-timestamp to the time of the last changelog entry if
|  | the maintainer does not set builttimeStamp explicitly.
|  | With this change almost all GNU R packages in Debian would automatically
|  | build reproducible.
|  
|  And I presume that is what we want?  I never heard from the reproducibility
|  group after I made the two patches (cf old bug report) to R itself and
|  r-cran.mk.
| 
| Without having further background information I would say that the build-
| timestamp should be either set to the build time (to carry correct 
| information) or not at all (to have reproducible builds).

Context: R _always_ puts one hin, hence the discussion in the earlier bug
report and my (long-accepted) upstream ptahc to be able to override.

What Philip is suggesting here is to use the hook I provided with a suitable
value -- the timestamp from debian/changelog.  I think I'll do that.

Dirk, at a workshop
 
| Johannes
| 
|  Too bad you sent me this today. I just made four uploads of R in the last
|  few days leading up to R 3.2.0.
|  
|  Dirk
|  
|  | Best,
|  | Philip
|  | 
|  | x[DELETED ATTACHMENT usable_builttimeStamp2.patch, text/x-patch]
|  | x[DELETED ATTACHMENT signature.asc, application/pgp-signature]

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-18 Thread Philip Rinn
Hi,

I think this patch is more what I wanted.
It now sets the build-timestamp to the time of the last changelog entry if the
maintainer does not set builttimeStamp explicitly.
With this change almost all GNU R packages in Debian would automatically build
reproducible.

Best,
Philip

--- r-cran.mk.orig	2015-04-18 11:05:30.874073000 +0200
+++ r-cran.mk	2015-04-18 11:10:04.116593671 +0200
@@ -68,16 +68,12 @@
 ## optional installation of a lintian silencer
 lintiandir	:= $(CURDIR)/debian/$(package)/usr/share/lintian/overrides
 
-## set built-time in DESCRIPTION time of created binary package based on stamp in changelog
-## cf discussion in http://bugs.debian.org/774031 --- and uncomment two assignments here
-##
-## extract built-timestamp from entry changelog and use as argument 
-#builttime   := $(shell dpkg-parsechangelog -l$(CURDIR)/debian/changelog | awk -F': ' '/Date/ {print $$2}')
-##
-#builttimeStamp  := --built-timestamp=\$(builttime)\
-##
-## else
-builttimeStamp  := 
+## if no builttimeStamp is known set built-time in DESCRIPTION to time of created
+## binary package based on stamp in changelog. Cf discussion in http://bugs.debian.org/774031
+ifeq ($(builttimeStamp),)
+  builttime   := $(shell dpkg-parsechangelog -l$(CURDIR)/debian/changelog | awk -F': ' '/Date/ {print $$2}')
+  builttimeStamp  := --built-timestamp=\$(builttime)\
+endif
 
 common-install-indep:: R_any_arch
 common-install-arch:: R_any_arch


signature.asc
Description: OpenPGP digital signature


Bug#782764: r-base-dev: Make it possible for maintainers to set builttimeStamp in d/rules

2015-04-17 Thread Philip Rinn
Package: r-base-dev
Version: 3.1.1-1
Severity: normal
Tags: patch

Hi,

as GNU R 3.2.0 is now released with support for setting the build timestamps it
would be nice to be able to use it.

You could just uncomment lines 75-77 in r-cran.mk to make that happen or leave
the decision to the maintainer and apply my proposed patch.

Maintainers would just have to set builttimeStamp in d/rules:
builttime   := $(shell dpkg-parsechangelog -l$(CURDIR)/debian/changelog |
awk -F': ' '/Date/ {print $$2}')
builttimeStamp  := --built-timestamp=\$(builttime)\


Best,
Philip
--- r-cran.mk.orig	2015-04-17 16:26:02.176622000 +0200
+++ r-cran.mk	2015-04-17 16:26:55.241509934 +0200
@@ -77,7 +77,9 @@
 #builttimeStamp  := --built-timestamp=\$(builttime)\
 ##
 ## else
-builttimeStamp  := 
+ifeq ($(builttimeStamp),)
+  builttimeStamp := 
+endif
 
 common-install-indep:: R_any_arch
 common-install-arch:: R_any_arch