On Mon, 2019-01-28 at 08:09 +0800, hagar wrote:
On 28/1/19 7:57 am, Mark Weiman wrote:
On 27/1/19 11:27 pm, Eli Schwartz via aur-general wrote:
On 1/27/19 6:13 AM, Hagar wrote:
Hello,
Can someone please have a look on what happens on server side?
I want to update gtkhtml4 in AUR adding a long proposed (in the comments) patch.
This fails with
fatal: unable to access ' https://aur.archlinux.org/gtkhtml4.git/': The requested URL returned error: 403
I maybe do not see the obvious, but I can update my other packages without problems, but not thi one, which I newly adopted.
Best Regards
Stefan Its not well documented - but edit the .git/config file an change
On 27/1/19 7:01 pm, stefan-husmann@t-online.de wrote: the url from https to ssh:
look at the aur page for the package for the exact url.
I hit this problem just last week. It is pretty well documented.
- We repeatedly document the use of ssh cloning everywhere in the wiki page describing the submission process, and make no mention of using https://
- When logged into the AUR and viewing a package that you maintain, it lists two clone urls: https:// and ssh:// -- and right after the https:// link it specifies in parentheses, "read-only". Read- only means you cannot write to it.
- When viewing a package that you do not maintain or when not even logged into the AUR, only the https:// clone url is referenced, and it still states "read-only".
- Other websites which support pushing over https:// will require you to type in your username and password every time you do, which is unfriendly and I don't understand why anyone would ever want to do so in the first place if they could just use ssh.
True to an extent.
You never specifically say to use ssh to check out the package.
It seems to be just implied. Your examples use both https and ssh.
The problem lies with the section - Acquire build files.
in this section you use http://...
Then later on the instructions for publishing make there is no mention of checking the .git.config to ensure that the correct url is in use.
That simple bit of information is missing.
Those of us who are unfamiliar with git need to know these simple things.
It would be nice if it was specifically mentioned in the instructions -eg.
Please check the .git/.config file of your repository to ensure that your url
is of the form ssh://... .If you get a 403 error you have the wrong url configured. A 403 does not mean you have an incorrect URL. A 403 means the server understood the request and it exists, but you do not have
On Mon, 2019-01-28 at 07:42 +0800, hagar wrote: permission to use that resource. It is rather clear that the https URLs are read only on the AUR and a 403 error is the correct error to state that you cannot push over that protocol as you made a request that you do not have permission to do.
The AUR section you're talking about that mentions acquiring build files is just that, for building. Not pushing updates to the PKGBUILD. It is understood that if you read the wiki and follow the directions on how to publish changes to the AUR, you use the SSH protocol and if you see a 403 error on HTTP if you make a mistake and use that, that you will be capable to interpreting that as you made a mistake.
Just my $0.02.
Mark
Thank you for clearing that up - I found no reference to what a 403 error actually was.
All I wish is for the editing of .git.config to be documented with the note to use the ssh://
url to checkout if you wish to push changes back.
This was the first time I had used git for this purpose and had to debug it the hard way.
As I mentioned a couple of months back - Not everyone has the same communication skills.
What may be clear to you may not be to someone else. I learned the hard way that -
"There is no such thing as common sense."
All I wish to do is help improve the AUR.
Thanks
Macca
HTTP 403 is an HTTP code and is therefore documented with the HTTP RPC. .git/config is a file part of a git repository and is documented with the git application. The use of git especially is understood to be, well, understood before using the AUR. Arch Linux, as far as I'm aware, assumes you know about how to use tools like git and I would also assume that you are capable of figuring out what an HTTP error code means based on something simple like the Wikipedia page that discusses HTTP status codes in detail. If you feel otherwise though, the Arch Wiki can be edited by any Arch user that creates an account on the Wiki, but that's not something the TUs or others like me in this channel are responsible for since its kinda assumed to be at least familiar with git before using it (which the AUR uses it). Mark