I've noticed for a while now, from the US, cloning "git.openwrt.org/openwrt/openwrt.git " has been very slow. Sometimes the connection will die and other times it will crawl at 200kbps or less.
I've tried from various locations in the US and the same result. I've also tried from European locations and it is a little better, but not by much.
I'm not sure what is causing the slow down, but wanted everyone to be aware.
1 Like
slh
August 4, 2019, 7:44pm
2
That seems to be peering related, from two locations (residential VDSL2 and data centre) in Germany I'm getting the maximum of my line speed:
$ time ( LANG= git clone https://git.openwrt.org/openwrt/openwrt.git/ )
Cloning into 'openwrt'...
remote: Enumerating objects: 473368, done.
remote: Counting objects: 100% (473368/473368), done.
remote: Compressing objects: 100% (129145/129145), done.
remote: Total 473368 (delta 327677), reused 468385 (delta 324197)
Receiving objects: 100% (473368/473368), 147.39 MiB | 15.12 MiB/s, done.
Resolving deltas: 100% (327677/327677), done.
real 0m27,094s
user 0m29,249s
sys 0m4,872s
diizzy
August 4, 2019, 8:08pm
3
GitHub works fine otherwise...
A little different story here in the States. This download attempt completely failed.
davidc502@Ryzen-3900x:~/Desktop$ time ( LANG= git clone https://git.openwrt.org/openwrt/openwrt.git/ )
Cloning into 'openwrt'...
remote: Enumerating objects: 473368, done.
remote: Counting objects: 100% (473368/473368), done.
remote: Compressing objects: 100% (129145/129145), done.
error: RPC failed; curl 56 GnuTLS recv error (-9): A TLS packet with unexpected length was received.
fatal: The remote end hung up unexpectedly
fatal: early EOF
fatal: index-pack failed
real 7m19.203s
user 0m0.001s
sys 0m0.011s
@diizzy
Yeah, a little bit different story when cloning into GitHub. I guess I can just do it that way. Thanks
davidc502@Ryzen-3900x:~/Desktop$ time ( git clone https://github.com/lede-project/source.git openwrt1 )
Cloning into 'openwrt1'...
remote: Enumerating objects: 159, done.
remote: Counting objects: 100% (159/159), done.
remote: Compressing objects: 100% (127/127), done.
remote: Total 473414 (delta 61), reused 82 (delta 27), pack-reused 473255
Receiving objects: 100% (473414/473414), 167.04 MiB | 46.70 MiB/s, done.
Resolving deltas: 100% (321955/321955), done.
real 0m7.018s
user 0m9.856s
sys 0m1.054s
diizzy
August 4, 2019, 8:59pm
6
Please use the "proper" url instead
https://github.com/openwrt/openwrt --> https://github.com/openwrt/openwrt.git
You can also use git:// instead of https
1 Like
@diizzy
sha256sums show the 2 repositories are different. However, diff didn't find a difference.
Interesting
Same/similar to what I was experiencing
Lately i've been getting lots of timeouts and retries like:
curl: (18) transfer closed with 46504538 bytes remaining to read
This is when attempting to download linux-4.14.131.tar.xz from https://downloads.openwrt.org/sources/linux-4.14.131.tar.xz using the build process.
It happens with other packages as well, although they usually are sufficiently small that they manage to crawl through the pipe to completion.
Am i alone in this or has anyone else seen it?
Geographically in Australia if iā¦
Are your downloads of the package sources also slow?
Same issue... This has been going on for some time. I haven't said anything because these things usually clear up over time. However, this issue has been progressively getting worse.
diizzy
August 4, 2019, 11:26pm
10
Bump: Is it too much asked for, to update the docs in case of (serious) changes ?
BTW, it has the advantage of less questions here on the forum.
May be, I am a too oldfashioned prgrammer. As almost half a century ago, it was good practice for us, to keep the docs up-to-date.
tmomas
September 1, 2019, 4:42pm
13
Yeah, good old days back then
Today, as a rule of thumb for open source: If you want something done, do it yourself.
Regarding the wiki: It's a wiki (ORLY? ), you can correct mistakes yourself.
Not to start a general, new topic about coding standards, I might first check, whether such open source practice, resulting in non-deterministic software, is already covered in an article in ACMs (Association of Computing Machinery) "Risk to the Public".
As documentation of software functionality by a user, instead from the author, is seriously at risk being wrong or incomplete.
Which is in strong contrast to the "security" fans around.
2 Likes
jow
September 2, 2019, 7:32am
15
Which serious changes? The git.openwrt.org domain is still the canonical git clone url.
but often works terribly slow. github.com/openwrt/openwrt.git much faster.
jow
September 2, 2019, 1:54pm
17
Which is besides the point of your message claiming that there is undocumented serious changes to the git cloning process.
Then what does this mean ?
jow
September 2, 2019, 2:43pm
19
I guess it means that @diizzy personally prefers to use the Github source mirror.
Which means, it is a good idea, to update the wiki like
"The official link https://github.com/openwrt/openwrt quite often is very slow, or does not work at all. So you might use the Github source mirror instead, https://github.com/openwrt/openwrt.git " .