More tests failing due to case sensitivity with curl 8.20.0

[charles2910]

Hi,

We are seeing 3 more tests failing with curl 8.20.0 in Debian: https://ci.debian.net/packages/t/trurl/testing/amd64/70698738/#L244

 61s 52: failed	--url https://curl.se/hello --set query=user=me
 61s --- stdout --- 
 61s expected:
 61s 'https://curl.se/hello?user%3dme\n'
 61s got:
 61s 'https://curl.se/hello?user%3Dme\n'
[...]
 61s 132: failed	-s scheme:=http -s host:=localhost -s path:=/ABC%5C%5C -s 'query:=a&b&a%26b'
 61s --- stdout --- 
 61s expected:
 61s 'http://localhost/ABC%5c%5c?a&b&a%26b\n'
 61s got:
 61s 'http://localhost/ABC%5C%5C?a&b&a%26b\n'
[...]
 61s 72: failed	https://example.com/one/t%61o/%2F%42/ --append path=%61
 61s --- stdout --- 
 61s expected:
 61s 'https://example.com/one/tao/%2fB/%2561\n'
 61s got:
 61s 'https://example.com/one/tao/%2FB/%2561\n'

I've prepared a patch (though it's a bit painful to have duplicated tests only with hex case difference :-( ) and will open a merge request.

And since we are here, could we have a 0.17.0 release? :p

[vszakats]

Seen also in trurl's curl-for-win CI jobs:
#439 (comment)

Thanks for reporting and fixing!

[charles2910]

#441 doesn't cover all cases... #395 introduced the uppercase-hex keyword, the problem is it triggers for curl 8.15 or newer. But these 3 tests only fails with 8.20 and newer. So if you run the tests against curl 8.19.0 the tests will fail too as seen in https://github.com/curl/trurl/actions/runs/25233678819/job/73994953019?pr=441#step:6:9

We'd need a more granular keywords to handle things between 8.15 and 8.20 in a sane way

[diogotcorreia]

6 more tests seem to be failing with curl 8.21.0:

34: failed	--url 'https://example.com/#%2e%61%13%Fa'
--- stdout --- 
expected:
'https://example.com/#.a%13%fa\n'
got:
'https://example.com/#.a%13%FA\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0
[...]
173: failed	https://example.com/one/t%61o/%2F%42/
--- stdout --- 
expected:
'https://example.com/one/tao/%2fB/\n'
got:
'https://example.com/one/tao/%2FB/\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0
[...]
186: failed	'http://example.com/?a=%5D'
--- stdout --- 
expected:
'http://example.com/?a=%5d\n'
got:
'http://example.com/?a=%5D\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0

187: failed	'http://example.com/?a=%5D&b=%5D'
--- stdout --- 
expected:
'http://example.com/?a=%5d&b=%5d\n'
got:
'http://example.com/?a=%5D&b=%5D\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0

188: failed	'sftp://us%65r:pwd;giraffe@odd'
--- stdout --- 
expected:
'sftp://user:pwd%3bgiraffe@odd/\n'
got:
'sftp://user:pwd%3Bgiraffe@odd/\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0
[...]
191: failed	'sftp://us%65r:pwd;giraffe@odd' --get '[:password]'
--- stdout --- 
expected:
'pwd%3bgiraffe\n'
got:
'pwd%3Bgiraffe\n'
--- stderr --- 
expected:
''
got:
''
--- returncode --- 
expected:
0
got:
0

[charles2910]

Indeed there are more tests failing. @vszakats I started to look and I think it won't be that difficult to create a python function to run before the string comparison and make all percent-encoded upper case, hopefully I'll have a patch soon (tm)