urljoin - broken links, duplication of the last base url segment

[alexbelyaev]

It looks like there is an issue in urljoin function.
When a relative URL is appended to a base URL, it duplicate the last segment of the base path in this case.

Base url + relative href = broken link with duplication
https://www.jobs-oberlausitz.de/stelle/ + stelle/detail/36993 =
https://www.jobs-oberlausitz.de/stelle/stelle/detail/36993
Correct:
https://www.jobs-oberlausitz.de/stelle/detail/36993

To Reproduce

CSS Selector bridge

URL:
https://www.jobs-oberlausitz.de/stelle/suche?filter%5Bregion_id%5D%5B0%5D=1&filter%5Btyp%5D%5B0%5D=Vollzeit&filter%5Bmerkmal_id%5D%5B0%5D=&Freitextsuche=&orderby=Datum&seitennummer=1

Article link selector (<a href="stelle/detail/37574" ... ):
#siteobjects-middle > form > div > div.siteobject-stellenliste > table > tbody > tr

Resulting links:

https://www.jobs-oberlausitz.de/stelle/stelle/detail/37574
https://www.jobs-oberlausitz.de/stelle/fileman/imgsc/fitheight/50/firmen/logo_64_St_dtisches_Klinikum_G_rlitz_gGmbH.png

Or:

urljoin( 'https://www.jobs-oberlausitz.de/stelle/', 'stelle/detail/36993')

Expected behavior

Result links:

https://www.jobs-oberlausitz.de/stelle/detail/37574
https://www.jobs-oberlausitz.de/fileman/imgsc/fitheight/50/firmen/logo_64_St_dtisches_Klinikum_G_rlitz_gGmbH.png

Desktop (please complete the following information):
Win 10
Chrome Version 131.0.6778.205 (Official Build) (64-bit)

[dvikan]

the first example looks correct to me. The relative part is appended to base.

[alexbelyaev]

But in the browser it works differently in fact, so the result link is broken in such scenario even though it works on the original page (in Chrome). Here is part of GPT explanation: "When a relative URL is appended to a base URL, it does not duplicate the last segment of the base path. Instead, it appends the relative URL directly to the base URL's directory." I've added GPT's version of the library for studying purpose.
urljoin.zip

[dvikan]

base: http://example.com/foo/bar/
relative: bar

Do you think it's correct to resolve this to http://example.com/foo/bar/ ?

See https://github.com/fluffy-critter/php-urljoin/blob/main/tests/cases.json for test suite

[alexbelyaev]

It is above my understanding why on the provided page it works the way it works (and how it should). But the result link in rss-bridge is different from what I have in the browser on that page. I assumed that the library was not taking something into account.

[dvikan]

The base url is https://www.jobs-oberlausitz.de/ as specified in the html tag <base>.

Still might be a bug here but it's not in urljoin. I think bug is that rssbridges does not respect the <base> tag in html pages.

[dvikan]

@ORelio @LarsStegman