Recompilation problems opencv (python3-cv2 --> python3.8)

Brian Inglis Brian.Inglis@SystematicSw.ab.ca
Fri Jan 15 17:14:57 GMT 2021


On 2021-01-15 01:13, wilhelm-eger--- via Cygwin wrote:
> Gesendet: Donnerstag, 14. Januar 2021 um 23:52 Uhr
> 
>> Please try by downloading the complete Cygwin source package, and try a vanilla
>> rebuild first with zero changes, then change only the .cygport VERSION= to the
>> latest, rebuild, and fix any problems with that as you go.
> 
> I have done the following
> 
> mkdir opencv
> cd opencv
> git clone http://cygwin.com/git/cygwin-packages/opencv.git .
> cygport opencv.cygport fetch
> cygport opencv.cygport all
> 
> Same error.
> 
> Is there any other location where I could get the opencv source package from?

Check your email host location:

$ geoiplookup web.de
GeoIP Country Edition: DE, Germany
GeoIP City Edition, Rev 1: DE, 01, Baden-Wurttemberg, Karlsruhe, 76137, 
49.001900, 8.428700, 0, 0
GeoIP ASNum Edition: AS8560 1&1 Internet SE

Check Cygwin German mirror locations for one nearby:

$ sed '/^\s\+[^;]\+;\([^;]\+\);Europe;Germany$/!d;s!!\1!' /etc/setup/setup.rc | 
sort -u | xargs -l1 geoiplookup
...
$ geoiplookup ftp-stud.hs-esslingen.de
GeoIP Country Edition: DE, Germany
GeoIP City Edition, Rev 1: DE, 01, Baden-Wurttemberg, Stuttgart, 70173, 
48.766701, 9.183300, 0, 0
GeoIP ASNum Edition: AS553 Universitaet Stuttgart
...

in same Land, lat/long diff about .25/.75, and about 61km great circle distance.

Find mirror URLs:

$ fgrep esslingen.de /etc/setup/setup.rc
 
ftp://ftp-stud.hs-esslingen.de/pub/Mirrors/sources.redhat.com/cygwin/;ftp-stud.hs-esslingen.de;Europe;Germany
 
http://ftp-stud.hs-esslingen.de/pub/Mirrors/sources.redhat.com/cygwin/;ftp-stud.hs-esslingen.de;Europe;Germany
 
https://ftp-stud.hs-esslingen.de/pub/Mirrors/sources.redhat.com/cygwin/;ftp-stud.hs-esslingen.de;Europe;Germany

Find source package:

$ grep -m1 '^source:\s.*opencv' ~/mirror/x86_64/setup.ini
source: x86_64/release/opencv/opencv-3.4.1-2-src.tar.xz 138322784 
8da273070a232773d522dfe9b2cf098115e34a595a3d0c968b7fc0ba4f17f92db4f405857157f72f5218e17e6694c1d93ddc22c70fc6bcbf383077499373657a

Download source package:

$ curl -OJRSs 
https://ftp-stud.hs-esslingen.de/pub/Mirrors/sources.redhat.com/cygwin/x86_64/release/opencv/opencv-3.4.1-2-src.tar.xz
*OR*
$ wget -N -nv 
https://ftp-stud.hs-esslingen.de/pub/Mirrors/sources.redhat.com/cygwin/x86_64/release/opencv/opencv-3.4.1-2-src.tar.xz

$ tar -xf opencv-3.4.1-2-src.tar.xz

Also package *patches* visible from:

https://build.opensuse.org/package/show/openSUSE:Factory/opencv3

but patches directly downloadbale from:

https://github.com/bmwiedemann/openSUSE/tree/20201007/packages/o/opencv3

[thanks to Achim for pointers]

and

https://sources.debian.org/data/main/o/opencv/3.2.0+dfsg-6/debian/patches/
https://salsa.debian.org/science-team/opencv/-/tree/debian/3.2.0+dfsg-6/debian/patches
https://salsa.debian.org/science-team/opencv/-/tree/debian/3.3.0+dfsg-1_exp0/debian/patches
https://salsa.debian.org/science-team/opencv/-/tree/debian/3.4.4+dfsg-1_exp1/debian/patches
https://salsa.debian.org/science-team/opencv/-/tree/debian/3.4.5+dfsg-1_exp1/debian/patches

and

https://src.fedoraproject.org/rpms/opencv/tree/f30

and you can pick later branches or tags for patches for later versions as you 
need from these repos: cygport can use URLs or local file names in 
PATCH_URI="..." and the string content may be a separate line per patch, usually 
with a leading tab.

> Btw. there is no need to change to another Version, I just want to have cv2
> also in python3.8 working. From what I have experienced with this package I
> assume that this would also not compile with python3.6 atm. Seems to be a
> cmake or gcc error to me.
First check your current packages, toolchains, and libraries against a known 
good working set of vanilla sources at an earlier version that should have no 
issues, then work your way up to newer versions adding difficulties! ;^>

-- 
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada

This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in binary units and prefixes, physical quantities in SI.]


More information about the Cygwin mailing list