<aparcar[m]>
did you delete ./tmp? maybe there is some magic
<mangix>
no
<mangix>
replace $(eval $(call SetupHostCommand,diff,Please install GNU diffutils, \ with $(eval $(call SetupHostCommand,gnu-diff,Please install GNU diffutils, \
<mangix>
this is quite curious...
<mangix>
it means this check has been broken
<aparcar[m]>
yea I get it
tohojo has joined #openwrt-devel
<mangix>
i tried replacing SetupHostCommand with TestHostCommand but that didn't work
<mangix>
pushed a fixed version anyway.
<mangix>
Any other issues with the prereq system someone else can deal with
<aparcar[m]>
so why do we need gnu-diff?
tohojo has quit [Ping timeout: 480 seconds]
<mangix>
my commit didn't explain?
tohojo has joined #openwrt-devel
<aparcar[m]>
i'll look again
<aparcar[m]>
mangix: no, you don't state why it need the rename from diff to gnu-diff
<aparcar[m]>
below is cp, I guess that doesn't need to be gnu-cp right?
<mangix>
I mean....that's Makefile arcanery that I'm not willing to unravel. All I know is, it doesn't fail even though it should without renaming diff to gnu-diff
<mangix>
aparcar[m]: turns out, all of them are broken
<mangix>
SetupHostCommand does not work properly.
<aparcar[m]>
mangix: PRs welcome ;)
<aparcar[m]>
that must be a recent thing
<aparcar[m]>
I remember it failing multiple times with missing tools
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<mangix>
actually yeah I remember now. There was discussion on replacing quilt's diff utility. Turns out it's hardcoded
<aparcar[m]>
mangix: so you're not planing to patch that but rather require GNU diff? tsts
<mangix>
...
<mangix>
that is way too intrusive
<mangix>
especially for a core utility like quilt
<slh>
the only reasonable way would be to get such a change included into upstream quilt
<mangix>
people have tried. nothing applied upstream so far.
<mangix>
in any case, quilt is responsible for make package/x/refresh . I'm not willing to potentially break that.
<slh>
once git am would be used internally by OpenWrt, binary patches (or just executable file permissions) would follow quite naturally, in which case all hell would break lose on developer's systems - where, at least for interatice usage, the system provided quilt binary wouldn't understand its syntax
<mangix>
I feel like a lot of these tools/ should be removed
<aparcar[m]>
merged
<aparcar[m]>
mangix: which and why?
<mangix>
aparcar[m]: patch for example. there's already a requirement on host patch
<mangix>
i really don't want WIP patches to the kernel or an out of tree module
<slh>
especially not for ntfs3
<slh>
exfat was kind of different in that regard (initially little hope for getting it merged - and not touching system partitions, so less of a risk)
<mangix>
speaking of exfat
<mangix>
maybe the out of tree module should be brought back for 21.02
<mangix>
I brought it in the took it out for a silly reason
<mangix>
/s/the/and
<mangix>
swalker: taglib info is wrong on uscan.
<slh>
didn't exfat initially get merged into v5.4, if so the easiest way is to stick to the mainline implementation and to let -stable handle potential issues. it's still better than the old out-of-tree module, and for anything else --> v.510
<mangix>
slh: it got merged with 5.7 IIRC
<mangix>
5.4 has a low quality version in the staging dir
<slh>
ah, yep - I remember... should be still roughly on par with the out-of-tree version we had before
ylk has joined #openwrt-devel
<mangix>
slh: not according to the samsung employee who maintains it