On Monday 27 April 2009 19:53:15 Peter Tyser wrote:
On Mon, 2009-04-27 at 19:41 -0400, Mike Frysinger wrote:
On Monday 27 April 2009 19:18:21 Wolfgang Denk wrote:
In message Peter Tyser wrote:
<base64 stuff>
...
Please do not post base64 encoded messages. Next time I will ignore such patches.
it's due to unicode yet again:
bmp_logo.c: In function `main':
those ticks are probably the left & right unicode ticks that pretty much all GNU tools use. it's pretty hard for people to catch these things in the normal flow of git usage considering using a unicode locale setting and a GNU system is by and far the norm.
Ahhhh, thanks for the enlightenment - that makes sense. I use urxvt and pasted the output of a gcc warning in the commit message.
which is reasonable and common behavior imo
The copies cc-ed to myself come through just fine as utf-8 fwiw. Does that imply the denx.de servers convert unicode messages to base64?
`git format-patch` did the right thing and converted the e-mail body to base64. if you do the format-patch again and edit the resulting file, you'll see it.
In general, what's the proper workaround so things like this don't happen?
there isnt a good one which is why i keep whining at Wolfgang to let it through. you can either not use a unicode environment at all (which sucks), or just remember this kind of thing and always check the output of git's format-patch before sending it. then when you notice, review the files to figure out where the unicode is coming from -- most of the time it's the commit message, but sometimes it comes from comments in the source code. so you have to keep your eyes open for umlauts and such (ëïöü).
i havent checked, but using extended codepages will probably also result in base64 encoding ... -mike