And yes you could make a regex that detect any sets of 8 spaces before
non-whitespace text, but that seems like a lot of effort for no good
reason, when not dealing with Makefiles.
Anything that applies a *hard* rule about whitespace is broken -
whitespace shouldn't be getting converted unless it's actually
necessary; a hard rule will fail under some conditions, and the real
issue is editors *changing* whitespace without user interaction (which
causes patches to fail).
Indead git's insistence on the exact same whitespace is the root issue,
since changing git would be a lot easier than changing a bazillion
editor to suite git (hint: ain't gonna happen).
I'm sure this has been discussed to death, but having patchwork be a
knob isn't the answer.
Regards,
Daniel
On 10/01/16 05:20 AM, Daniel Dickinson wrote:
Hi,
On 10/01/16 05:10 AM, Gert Doering wrote:
Hi,
On Sun, Jan 10, 2016 at 04:47:05AM -0500, Daniel Dickinson wrote:
I used git send-email so there shouldn't be any whitepspace mangling
issues, unless patchwork is to blame.
Over at openvpn-devel, we recently discovered that some versions of
MS Exchange mangle whitespace for "mails in transit" - so even when the
user did everything right (git send-email, etc.) Exchange still converted
all tabs to 8x Space, leading to non-applying patches.
I think this is a different issue. The issue here is that there is
indeed 8 spaces instead of a tab, *however* it's in a shell script AND
git doesn't complain, nor do regexes that detect things like spaces
before tabs (there is no tab after the spaces, only non-whitespace
text). In short patchwork appears to say 'for *all* files, you *must*
use tab instead of 8 spaces', which is not reasonable, and in some cases
is actually wrong.
Also I am pretty certain provider is not running Windows, nor is
OpenWrt, so that is not likely to at issue.
Solution: send via freemail provider, using direct SMTP+Auth (which
git send-email can do).
That leaves receive-side error. It seems likely given the line the
error is on, that patchwork is pickier than git or the vim bit you sent
me. That is the 'offending' line is a shell script that happens to have
8 spaces instead of tab. Methinks this is not a relevant thing to
complain about in a shell script; also I wonder whether it would do the
same with with e.g. Python? I suspect patchwork is applying Makefile
rules to all patches, which is just wrong.
The problem often is context. If the parts before or after the actual
change get whitespace-modified, git am will not accept it (traditional
patch will, and complain about fuzz needed).
However, the error being reported is not 'patch does not apply', it is
'whitespace error', which implies that it's something patchwork is
reporting.
Perhaps the error report really should be 'patch does not apply because
of whitespace error', but that is not what I was told, so I'm left
guessing.
Regards,
Daniel
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
