From: Segher B. <se...@ke...> - 2011-01-31 15:31:10
|
>> No, you even quoted the question. Segher asked if there's any problem >> not >> having ANY attribute. You're answering about whether eol=lf and >> eolf=crlf >> works. So, if you put it in the repo with no attributes and with >> autocrlf=true, don't you get a CRLF working copy? I thought that was >> what >> we used to do. > > Ah, OK. Then there are 2 answers there: > - with autocrlf=true, not that I know of > - without autocrlf=false, then obviously yes But then all other text files will not have the correct line endings either for your toolchain (they will be LF instead of CRLF). So, in all cases, it works fine to have it as a text file without anything else? > 1. people should be free to set their options to whatever they like, if > we can support it at no cost And the manweeks spent on this stupid discussion are "no cost"? > So we're pretty much trying to argue "But git or diff might be buggy > with regards to CRLF handling!". If that's the line you really want to > take fine, but we might as well advertise "libusb might be buggy - > better avoid it if you can". That is the line _you_ take, with the eol attribute handling. > The pragmatic approach (as well as the most convenient approach for our > git users, since it doesn't touch any of the files that UNIX/Mac people > would normally use) is to use -text and CRLF. You might not like it from > a standpoint of "being clean", but as far as I can tell, it is the > safest way to put this matter to rest and move on. So, does that work? What do patches look like on a unix system / in a mail client? Can such patched be applied on a unix system? Etc. Segher |