Forum > Other

Problems with Gitlab

(1/12) > >>

JuhaManninen:
Things that came up in the "Forum upgrade" thread.

I cannot login to Gitlab using my eMachines mini-laptop from year 2009. It has an Intel Atom CPU and 1GB RAM.
It has a modern MX Linux 23 OS, installed last summer, a little over a year ago.
Fifefox browser version is recent enough. Clearly the problem is not outdated SW in my laptop.
When trying to sign in, Gitlab wants to confirm I am a human. Usually there is a Cloudflare checkbox but the mini-laptop only shows an animated circle and asks me to check my internet connection.
All other sites can be opened with Firefox in this mini-laptop, although many of them make the browser hog memory and the system starts to swap.
Internet pages are getting heavier and heavier. 10 years ago internet was usable with the exact same laptop.
I blamed Javascript but is that the only reason? I don't even know about the new web technologies involved.

What more, I cannot login to Gitlab using a Falkon browser in my fast Ryzen computer. It has Manjaro Linux and latest versions of most SW.
This time the Cloudflare confirmation checkbox shows up. After clicking it the page shows an animated circle for some seconds and then the checkbox comes back. It repeats eternally.
Earlier I could make it work by setting a fake browser ID for only "gitlab.com" :
Mozilla/5.0 (Linux; Android 12) AppleWebKit/537.36 (KHTML, like Gecko) Version/4.0 Chrome/108.0.5359.79 DuckDuckGo/5 Safari/537.36
but it does not help any more.
Many sites now refuse to work with Falkon although it is a fully standards compliant browser. Many Google services and my e-mail service Tutanota.com are examples. The list grows. The claim that "a page works if Javascript is enabled" is just not true.

The advice for using a "light weight browser" to solve the resource problems does not work. A "light weight browser" is a myth. For example Falkon is light weight, integrates well with KDE and starts snappily. However when you open a heavy internet page, it hogs as much memory and is as slow as Firefox. There are many browsers but only a few rendering engines. Their speed and memory requirements are rather equal. The problem is the internet page itself!
@Aruna suggested using Lynx for Gitlab. Yeah, not a good joke...

Gitlab is super-slow even in a fast computer. I guess there is some fancy Java framework at the server side.
The old Mantis was made with PHP which is maybe the slowest programming language out there but still it was faster than Gitlab. :(

There is another irritation with Gitlab.
When I login from a different computer, it sends a verification code to my email and I must type it in. That would be OK if I signed in from a certain computer for the first time, but it happens always. If I switch between 2 computers I must use a verification code every time. I don't have my lazarus-ide.org mail configured everywhere which makes this super-annoying.
Can this be adjusted by project or is it built-in with Gitlab?

Summa summarum: This is a good example of Wirth's law. When computers get faster, people always find ways and excuses to make them slow again.

dbannon:
Hmm, I found earlier that my gitlab credentials no longer work. These credentials did once work and I have not changed the password. I can login (to gitlab) using github and my github credentials and that is what I usually do. So, I wonder if there is a time between use factor in play here ? Use it or lose it ?

Can you logon to gitlab in any other way ?

In terms of web pages (and their javascript) being resource hogs, I get much of my local news from our nation broadcaster, Australian Broadcasting Corporation but am now in the habit of shutting it down when finished as it will sometimes suck all available cpu ....

Davo

MarkMLl:
I wasn't able to progress into the sign-in screen using a laptop with Debian 5 ("Lenny") and (in effect) Firefox v3, apparently due to an SSL/TLS version incompatibility which the browser announced (i.e. there was not just a "blank screen" as insisted by at least one user) albeit with no specific version-negotiation trace. I think that in that particular case the connection was refused before it got to the point of trying to use Javascript, which means that any screen-size requirement (which would need Javascript detection) isn't relevant at this point (it might be later).

In this context I think https://gitlab.com/gitlab-org/gitlab-foss/-/issues/43433 is relevant, which I notice was closed without any suggestion that Gitlabs was interested in resolving it. However in fairness, I'd note that Github now has a similar Javascript requirement.

I don't know the extent to which the TLS protocol version can be forced or overridden, or even traced without using Wireshark or similar. Once the connection has been established something like https://hackaday.com/2022/03/22/wireshark-https-decryption/ might throw more light on what's going on.

Much if not all of this really boils down to inadequate error reporting: browsers not giving adequate information on what TLS version a server's infrastructure demands, a server not giving adequate information on what version of Javascript its software framework assumes (if anybody's actually worked that out) and so on.

MarkMLl

MarkMLl:

--- Quote from: dbannon on September 10, 2024, 11:28:34 am ---Hmm, I found earlier that my gitlab credentials no longer work. These credentials did once work and I have not changed the password. I can login (to gitlab) using github and my github credentials and that is what I usually do. So, I wonder if there is a time between use factor in play here ? Use it or lose it ?

--- End quote ---

I saw that a year or so ago: I can't remember the details but I think it was a one-off foulup by Gitlabs' management which was resolved after somebody at the FPC/Lazarus end prodded them.

I've got no specific record of my having to change a password etc., but I might possibly have had to recreate myself with the same userid and password.

MarkMLl

dbannon:

--- Quote from: MarkMLl on September 10, 2024, 11:37:46 am ---.....
However in fairness, I'd note that Github now has a similar Javascript requirement.
....

--- End quote ---

I did have a little play comparing github with gitlab before the previous thread was anonymously locked. (I am not a big fan of that, while I thoroughly respect an admin's right to step in, there should be a "who and why" statement. Please.

Anyway, back to Gitlab v Github ?  I have a U16.04 VM, I no longer test with it but its nice to know I could.  From that VM, and its Firefox 88, I could work with github (read only), it seemed fully functional.  Gitlab on the other hand would not open the Lazarus or FPC source repos in an way.

So, using my current system, turning Javascript off in Joanna's honour, github was usable, few things missing but usable. Gitlab, again, totally unusable.

As Oscar Wilde once said, "comparisons are odorous" but sometimes, they are revealing too.

Davo

Mark
(note, I did not put @MarkMLI)

> recreate myself with the same userid and password.

Wow, now, that is unproffesional !  I will continue to login via github.

Davo

Navigation

[0] Message Index

[#] Next page

Go to full version