Copy over pinnings from the webpage script #7

jakirkham · 2018-02-22T06:16:53Z

This is a mostly accurate copy of the content from the webpage pinning script to conda-forge-pinning.

ref: https://github.com/conda-forge/conda-forge.github.io/blob/61cdd7fd0374c07e95afdcbd414c7985b141e4b3/scripts/pin_the_slow_way.py

conda-forge-linter · 2018-02-22T06:16:56Z

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

jakirkham · 2018-02-22T06:17:52Z

recipe/conda_build_config.yaml

 bzip2:
  - 1.0
 cairo:
  - 1.14
 curl:
-  - 7.54
+  - 7.44


We seem to have an older pinning of curl.

jakirkham · 2018-02-22T06:18:53Z

recipe/conda_build_config.yaml

 bzip2:
  - 1.0
 cairo:
  - 1.14
 curl:
-  - 7.54
+  - 7.44
 dbus:
  - 1


This was not present in pin_the_slow_way.py. Given it seems pretty stable and has been at 1 a long time, I don't know if we need this, but I guess it can't hurt.

jakirkham · 2018-02-22T06:19:33Z

recipe/conda_build_config.yaml

@@ -70,27 +72,29 @@ flann:
 fontconfig:
  - 2.12
 freetype:
-  - 2.8                # [not ppc64le]
+  - 2.8.1              # [not ppc64le]
 gstreamer:
  - 1.12
 gst_plugins_base:
  - 1.12


These were missing from pin_the_slow_way.py. Though it is good to see them here.

jakirkham · 2018-02-22T06:21:04Z

recipe/conda_build_config.yaml

 geos:
  - 3.6.2
 giflib:
  - 5.1
 glib:
-  - 2.53
+  - 2.55


We are using glib 2.55 currently. Switching to 2.53 will undoubtedly break things based on various symbols added and one removed over time.

ref: https://abi-laboratory.pro/tracker/timeline/glib/

jakirkham · 2018-02-22T06:22:29Z

recipe/conda_build_config.yaml

 glpk:
-  - 4.62
+  - 4.61


My reading of the API/ABI report for glpk is that 4.61 and 4.62 are compatible, but 4.61 added one symbol after 4.60. Vaguely remember talking to @dougalsutherland about this some time ago.

Yeah, I think that's right. We should maybe just update to the new 4.65 (which has new symbols), though.

Could you please add that as a new issue? Would merely like to make sure we have copied everything over ok first. We can update them in a second pass.

Opened as issue ( #13 ).

jakirkham · 2018-02-22T06:25:01Z

recipe/conda_build_config.yaml

 libpng:
-  - 1.6
+  - 1.6.32


There was a rather strange constraint. Probably mostly based on our desire to use the same pinnings in build and run. This was the lower end of that constraint. So think this should be ok going forward, but we may need to take the high end of the constraint ( 1.6.34 ) if we want to be really conservative.

jakirkham · 2018-02-22T06:26:19Z

recipe/conda_build_config.yaml

 librdkafka:
-  - 0.9.5
+  - 0.9.4


Same story as glpk. Both 0.9.4 and 0.9.5 appear to be API/ABI compatible based on the report. So picked the low end of what we had in pin_the_slow_way.py.

jakirkham · 2018-02-22T06:27:37Z

recipe/conda_build_config.yaml

 libssh2:
  - 1.8
 libsvm:
  - 3.21
 libtiff:
-  - 4
+  - 4.0.8


Like libpng, there was a rather strange constraint. Probably mostly based on our desire to use the same pinnings in build and run. This was the lower end of that constraint. So think this should be ok going forward, but we may need to take the high end of the constraint ( 4.0.9 ) if we want to be really conservative. The latter matches defaults based on our last annotation in pin_the_slow_way.py.

jakirkham · 2018-02-22T06:30:32Z

recipe/conda_build_config.yaml

 r_base:
  - 3.3.2
  - 3.4.1
 scipy:
  - 0.19
 snappy:
-  - 1
+  - 1.1.6


Historically snappy is not that stable. At 1.1.5 symbols were added and removed plus the SONAME changed. Though 1.1.6 and 1.1.7 seem to be ok. Picked the lower end of what we had before again.

jakirkham · 2018-02-22T06:31:25Z

recipe/conda_build_config.yaml

 sox:
-  - 14.4
+  - 14.4.2


sox is really unstable even between patch versions. So this needs to be tight. Again same as what we had.

jakirkham · 2018-02-22T06:32:53Z

recipe/conda_build_config.yaml

 sqlite:
-  - 3
+  - 3.20


This is our pinning, but not what we are using. Please see issue ( conda-forge/sqlite-feedstock#25 ) for the publication issue. We may need to go to 3.21.0 to fix it.

jakirkham · 2018-02-22T06:33:36Z

recipe/conda_build_config.yaml

-xerces_c:
-  - 3
+  - 0.9.20
+xerces-c:


Not sure if having a - in the name is problematic.

it is. _ and - are equivalent in the config file

I mean _ in the config file and package name with - are equivalent

jakirkham · 2018-02-22T06:34:50Z

recipe/conda_build_config.yaml

-  - 3
+  - 0.9.20
+xerces-c:
+  - 3.2.0


This seems really tight, but that is what we and defaults have apparently. Did not find API/ABI info, but did request it in issue ( lvc/upstream-tracker#29 ).

Here's the newly added API/ABI report. Probably 3.2 is sufficient, but would be good to hear from xerces-c users. @conda-forge/xerces-c?

jakirkham · 2018-02-22T06:36:03Z

recipe/conda_build_config.yaml

 zeromq:
  - 4.2
 zlib:
-  - 1.2
+  - 1.2.11


zlib is kind of a mess. Pinning exactly seems safest.

jakirkham · 2018-02-22T06:39:11Z

There's some other stuff I skipped like zeromq where we currently have 4.2.1 due to mitigating circumstances. Not sure what we want to do with that.

This is a mostly accurate copy of the content from the webpage pinning script to `conda-forge-pinning`.

jakirkham · 2018-02-22T06:46:29Z

This copies over everything in pin_the_slow_way.py. We should stop using that script going forward and only update the pinnings in this feedstock. This will be used to automate pinnings in feedstocks going forward thanks to all the work done in conda-build 3 to make this possible and conda-smithy 3 to leverage it in the feedstocks.

cc @conda-forge/core

isuruf · 2018-02-22T06:49:08Z

Note that this alone will not be enough. All these packages need to rebuilt with run_exports set

isuruf · 2018-02-22T16:46:02Z

recipe/conda_build_config.yaml

 gmp:
-  - 6
+  - 6.1


6 is enough. Minor versions add symbols, but doesn't remove them. So, run_exports with a minor version as the min would fix this.

jakirkham · 2018-02-22T19:00:01Z

recipe/conda_build_config.yaml

@@ -167,7 +173,7 @@ pixman:
 poppler:
  - 0.61
 proj4:
-  - 4
+  - 4.9.3


Has changed SONAME in patch releases as noted in the ABI/API report. Seems good to keep this locked down pretty tight.

jakirkham · 2018-02-23T17:27:49Z

recipe/conda_build_config.yaml

@@ -167,7 +173,7 @@ pixman:
 poppler:
  - 0.61
 proj4:
-  - 4
+  - 4.9.3
 protobuf:
  - 3.4


We should bump this to 3.5. Also planning on making this libprotobuf as that will be the C++ only package. Will need to keep it at 3.5 for VS 2008 support as it is the last non-C++11 release, but will be a LTS release.

Opened as issue ( #14 ).

scopatz · 2018-02-23T18:36:05Z

recipe/conda_build_config.yaml

 zlib:
-  - 1.2
+  - 1.2.11


Will being more specific here break anything?

This is the current zlib pinning in the webpage repo. So no it shouldn't.

Some packages were implicitly using this version when they had the zlib 1.2.* pinning. In those cases, this pinning should fix breakage. Though ultimately those packages need to be rebuilt with a tighter pinning.

Also please see my old comment for details about ABI breakage between patch versions, which motivates this tight pinning.

jakirkham · 2018-02-27T19:10:08Z

Anything else here or is this good to go?

isuruf · 2018-02-27T19:13:34Z

recipe/conda_build_config.yaml

@@ -138,7 +144,7 @@ metis:
 mkl:
  - 2018
 mpfr:
-  - 3
+  - 3.1


3 is enough when use run_exports

djsutherland · 2018-02-27T21:18:10Z

recipe/conda_build_config.yaml

 glpk:
-  - 4.62
+  - 4.61
 gmp:
  - 6


pin_the_slow_way had 6.1.*; note that 6.1 added symbols. Maybe run_exports is enough though, I haven't really looked into how all of that works yet.

Please see this (outdated) comment. FWIU run_exports lower bounds the version based on the one used for the build. More details in this doc.

cc @isuruf

ENH patch libgfortran for v4

jakirkham commented Feb 22, 2018

View reviewed changes

jakirkham force-pushed the port_pin_the_slow_way branch from 2966f09 to 88128ec Compare February 22, 2018 06:41

jakirkham mentioned this pull request Feb 22, 2018

Updates #6

Merged

Copy over pinnings from the webpage script

2dce8fb

This is a mostly accurate copy of the content from the webpage pinning script to `conda-forge-pinning`.

jakirkham force-pushed the port_pin_the_slow_way branch from 88128ec to 2dce8fb Compare February 22, 2018 06:44

jakirkham mentioned this pull request Feb 22, 2018

Drops pin_the_slow_way for conda-forge-pinning conda-forge/conda-forge.github.io#534

Merged

isuruf reviewed Feb 22, 2018

View reviewed changes

jakirkham commented Feb 22, 2018

View reviewed changes

jakirkham commented Feb 23, 2018

View reviewed changes

scopatz reviewed Feb 23, 2018

View reviewed changes

Relax gmp pinning

5d3c8c7

isuruf reviewed Feb 27, 2018

View reviewed changes

Relax mpfr pinning

f7e1041

isuruf approved these changes Feb 27, 2018

View reviewed changes

scopatz approved these changes Feb 27, 2018

View reviewed changes

djsutherland reviewed Feb 27, 2018

View reviewed changes

isuruf merged commit 3308a5c into conda-forge:master Feb 28, 2018

isuruf mentioned this pull request Feb 28, 2018

Migrate latest pinnings from old pinning script #2

Closed

jakirkham deleted the port_pin_the_slow_way branch February 28, 2018 06:24

This was referenced Feb 28, 2018

Bump glpk to 4.65 #13

Closed

Bump protobuf to 3.5.0 #14

Closed

maresb pushed a commit to maresb/conda-forge-pinning-feedstock that referenced this pull request Nov 1, 2022

Merge pull request conda-forge#7 from beckermr/patch-1

44d542f

ENH patch libgfortran for v4

@@ @@ -138,7 +144,7 @@ metis: @@
               mkl:
                 - 2018
               mpfr:
-                - 3
+                - 3.1

Copy over pinnings from the webpage script #7

Copy over pinnings from the webpage script #7

Conversation

jakirkham commented Feb 22, 2018

conda-forge-linter commented Feb 22, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jakirkham commented Feb 22, 2018

jakirkham commented Feb 22, 2018

isuruf commented Feb 22, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jakirkham commented Feb 27, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

djsutherland Feb 27, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

djsutherland Feb 27, 2018 •

edited

Loading