Test/base64

[benjamin-si]

No description provided.

[benjamin-si]

Hey @andrew-svirin, this is tests for base64 error.

[andrew-svirin]

@benjamin-si there are some conflicts there.

[benjamin-si]

done

[andrew-svirin]

@benjamin-si

    PR #245 tests: flawed test design                                                     
                                                                                           
     All 3 "bug" tests (Test2/3/4) fail, but not for the reason the PR claims. The         
     tests assert assertEquals($original, $decoded) where $original = 1023 bytes, but      
     they feed in truncated base64 (1025-1027 chars sliced from a 1364-char string).       
     The data was chopped off before decoding — it's mathematically impossible to          
     recover 1023 bytes from that.                                                         
     ┌────────┬───────────┬────────────┬────────────┬─────────┬───────────────────────┐    
     │Test    │Input chars│decodeBuffer│base64_decod│Match?   │PR assertion vs        │    
     │        │           │            │e           │         │original               │    
     ├────────┼───────────┼────────────┼────────────┼─────────┼───────────────────────┤    
     │2 (1025)│1025       │768 bytes   │768 bytes   │Identical│FAIL (expects 1023)    │    
     ├────────┼───────────┼────────────┼────────────┼─────────┼───────────────────────┤    
     │3 (1026)│1026       │769 bytes   │769 bytes   │Identical│FAIL (expects 1023)    │    
     ├────────┼───────────┼────────────┼────────────┼─────────┼───────────────────────┤    
     │4 (1027)│1027       │770 bytes   │770 bytes   │Identical│FAIL (expects 1023)    │    
     └────────┴───────────┴────────────┴────────────┴─────────┴───────────────────────┘    
                                                                                           
     The critical column: decodeBuffer output is identical to base64_decode of the         
     full input. There is no data loss — the PR's assertEquals(original) is comparing      
     against the wrong expected value. 

     Real scenario: properly encoded unpadded base64                                       
                                                                                           
     All sizes pass — decodeBuffer matches base64_decode exactly, including remainders     
     of 2, 3, 343, and 523 bytes.                                                          
                                                                                           
     Verdict 

     The bug described in PR #245 does not exist. The tests are written incorrectly:       
     they truncate base64 data (destroying encoded bytes) then assert the original         
     data should be recovered. The decodeBuffer chunked decoder produces byte-             
     identical output to PHP's native base64_decode() on every input tested.

[andrew-svirin]

Found some buffering corner case. But it is with Zip. So you can also check if it somehow impact (3cdb8f2#diff-0c4af4208c073350c777161e27c65eb791835666a833247f04f6dc3cc4132aa1R123)

[benjamin-si]

Mhh, i will retry new test

[benjamin-si]

Last try for me.

I added testBuggyChunkedBase64DecodeCorruptsAesCiphertext

[benjamin-si]

I will fix it localy if you can't reproduce it

[andrew-svirin]

Maybe you have some PHP pedefined consts. Because localy tests from MR pass. Yes, if you can localy to fix, then make it locally.

[benjamin-si]

hey @andrew-svirin , updated to last version this morning is working now