Hi, I am trying to setup CircelCI for Windows and noticed an unexpected failure of some of the tests caused by modification of our binary files by CircleCI.
(Couldn’t find any information of that kind anywhere on the site, maybe due to incorrect keywords search, so logging.)
Compare results from the same job run on Linux and Windows.
Provided script checks out the code and prints hex contents of several binary files containing character 0a
as one of the first 2 bytes.
Content of these files appears to be different on Linux (expected) and Windows (unexpected).
Compare output from xxd hexcheck/10a.part
:
Linux:
00000000: 0a0a 0a0a 0a0a 0a0a 0a0a
Windows:
00000000: 0d0a 0d0a 0d0a 0d0a 0d0a 0d0a 0d0a 0d0a
00000010: 0d0a 0d0a
It seems like CircleCI exchanges 0a
with 0d 0a
newline characters when 0a
happens somewhere at the start of the file. (Exact logic is unclear. Substitution happens in all provided tests but “hex 3 bytes” and “hex 2 and 4”. Files where first 0a
happens on byte 3 or later seem not to be affected.)
Do not know if the same/similar issue would happen on MacOS.
This behavior was quite unexpected. Would it be possible to do something about it?
Thanks.