[OE-core] [PATCH 05/10] ncurses: fix incorrect UPSTREAM_CHECK_GITTAGREGEX

[Hongxu Jia]

On 6/1/19 4:27 AM, Alexander Kanavin wrote:
> On Sun, 12 May 2019 at 10:18, Hongxu Jia <hongxu.jia at windriver.com 
> <mailto:hongxu.jia at windriver.com>> wrote:
>
>     The upstream git tag has a `upstream/' prefix, such as:
>     >>> import re
>     >>> pattern = "upstream/(?P<pver>\d+(\.\d+)+(\+\d+)*)"
>     >>> string = "upstream/6.1+20181013"
>     >>> result = re.match(pattern, string)
>     >>> result['pver']
>     '6.1+20181013'
>
>
> Have you checked with devtool that this actually works?
>
I am afraid no, I did not run devtool it before


> Before this commit:
> alexander at alexander-box:~/development/poky$ devtool 
> check-upgrade-status ncurses
> INFO: ncurses                   6.1             6.1+20181013    Hongxu 
> Jia <hongxu.jia at windriver.com <mailto:hongxu.jia at windriver.com>> 
> 7a97a7f937762ba342d5b2fd7cd090885a809835
>
Then I know the script, and run it, I am afraid only revert is not 
sufficient.

The current ncurses version is `6.1+20181013' rather than `6.1', since 
upstream latest

version is `6.1+20181013', but auth email incorrect report ncurses 
should be upgraded

(that's why I try to fix UPSTREAM_CHECK_GITTAGREGEX, but it seems a 
wrong try)


Thanks to your mention of devtool, I know how auth works, and trace the 
root cause,

then I found in the following commit, it filters out `+' from git pv, 
but in ncurse case,

`+' is part of version

...

commit ab609c471d85be3248b789c8ab2813957cd97e29
Author: Alexander Kanavin <alexander.kanavin at linux.intel.com>
Date:   Mon Aug 3 19:29:54 2015 +0300

     recipeutils.py: allow all characters in regexes used to parse 
version strings

     Previously only numeric characters were matches and anything else was
     discarded, so 4.0-rc3, 2005e, 1.0.2a and similar versions got 
truncated.

     Signed-off-by: Alexander Kanavin <alexander.kanavin at linux.intel.com>
     Signed-off-by: Richard Purdie <richard.purdie at linuxfoundation.org>

diff --git a/meta/lib/oe/recipeutils.py b/meta/lib/oe/recipeutils.py
index 9d45b42..cd74213 100644
--- a/meta/lib/oe/recipeutils.py
+++ b/meta/lib/oe/recipeutils.py
@@ -638,7 +638,7 @@ def get_recipe_pv_without_srcpv(pv, uri_type):
      sfx = ''

      if uri_type == 'git':
-        git_regex = 
re.compile("(?P<pfx>(v|))(?P<ver>((\d+[\.\-_]*)+))(?P<sfx>(\+|)(git|)(r|)(AUTOINC|)(\+|))(?P<rev>.*)")
+        git_regex = 
re.compile("(?P<pfx>v?)(?P<ver>[^\+]*)((?P<sfx>\+(git)?r?(AUTOINC\+))(?P<rev>.*))?")
          m = git_regex.match(pv)

          if m:

...

I believe the commit is to parse manually  PV in **_git.bb (such as `PV 
= "1.9.0+git${SRCPV}"' ),

which contains `+gitAUTOINC+' in PV, but the regrex is not suitable for 
ncurses version


//Hongxu


> After the commit:
> alexander at alexander-box:~/development/poky$ devtool 
> check-upgrade-status ncurses
> INFO: ncurses                   6.1 UNKNOWN_BROKEN  Hongxu Jia 
> <hongxu.jia at windriver.com <mailto:hongxu.jia at windriver.com>>
>
> The code in bitbake/lib/bb/fetch2/git.py actually splits each tag with 
> '/' as separator and takes only the last part for some reason, so 
> those prefixes are never a part of the string that is matched against 
> regex.
> I'll send a revert in a moment.
>
> Alex
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20190603/a746889d/attachment.html>