13:06 <PaulFertser> So someone will have to package libjaylink for Homebrew.

13:27 <PaulFertser> Just noticed Jenkins is building mingw targets without capstone, hm.

14:37 <karlp> PaulFertser: one super tiny thing, for libjayink, on line 38 and line 45 you have different glob patterns for the library?

14:38 <karlp> is that intentional?

14:38 <karlp> the final install won't install a ".so" by itself?

14:38 <karlp> only a versioned one?

14:39 <karlp> a current owrt system has a weird mix, some with .so, .so.n and .so.n.xyz, and some with only .so.n and .so.n.xyz

14:39 <PaulFertser> borneoa_: build machine updated to provide libjaylink for all the targets.

14:40 <PaulFertser> karlp: good point, I guess I copied from hidapi assuming it's already sane. Guess so* should be used uniformly?

14:55 <karlp> I don't really know what's _right_ I just normally try and go for at least self consistent :)

14:55 <karlp> what does jaylink install build? does it build a .so symlink? or only the versioned ones?

14:56 <karlp> does openocd find the versioned one correctly?

14:56 <karlp> if it does, I'm ok with moving on with it :)

15:13 <PaulFertser> karlp: .so symlink is installed, yes. OpenOCD calls linker with -ljaylink so it's up to the linker to use the right name.

15:15 <PaulFertser> Updated, thank you again for the review.

15:34 <karlp> you really haven't run tested this? :)

15:38 <karlp> I can take both openocd and libjaylink,

15:38 <karlp> just going to build those real quick.

15:39 <karlp> I take it users who had openocd=y in their config with older openwrt are going to lose libjaylink now unless they manually select it right?

15:39 <karlp> that seems like the only right path, just checking I understand.

15:40 <karlp> you're doing a hard select on +libftdi1 and +hidapi and +libgpiod, but not libjaylink?

15:40 <karlp> I don't really mind what you want to do there, there's no good way of "suggests" in openwrt's world that I'm aware of :)

15:46 <karlp> woops, failed to compile for me.

15:47 <karlp> configure: error: ./configure.gnu failed for jimtcl

15:48 <karlp> this is after pulling your openocd 0.12 and libjaylink package updates, selecting openocd, and make package/openocd/{clean,compile}

15:57 <karlp> https://paste.centos.org/view/382a9383

15:57 <karlp> looks like jimtcl think's it's building for teh host or something?

15:59 <PaulFertser> karlp: no, current openocd 0.12.0 will build with internal libjaylink just fine.

15:59 <PaulFertser> karlp: but after libjaylink is added to repository I can add it to dependencies

15:59 <PaulFertser> karlp: and yes, indeed I haven't runtested it. No usecase, sorry...

16:00 <PaulFertser> Odd stuff, it builds on buildbots and builds for me locally.

16:02 <karlp> do you use ccache?

16:02 <PaulFertser> karlp: no

16:02 <PaulFertser> (I mean I use it Jenkins for OpenOCD but not for testing OpenWrt makefiles)

16:03 <karlp> it would be my first personal failure if it is this :)

16:03 <PaulFertser> I'd also check config.log to see how exactly it failed

16:05 <karlp> yep, it's busted with ccache. I'm going to say that's a "you" problem though, I'm not going to merge it when it doesn't build for me :)

16:06 <PaulFertser> borneoa_: btw, fascinating http://www.efton.sk/STM32/gotcha/index.html

16:07 <PaulFertser> karlp: hm, I've heard about ccache breaking things but I could never understand what specifally can lead to that.

16:08 <PaulFertser> Regarding packaging, once libjaylink is merged, I plan to send another patch to OpenOCD Makefile to +libjaylink to be future-proof and make it easier to build newer-than-0.12.0 versions.

16:10 <karlp> I mean, if you fix my ccache issue, I can just merge them together...

16:10 <karlp> with +libjaylink from the beginning.

16:11 <karlp> I'm just rebuilding it with ccache on again, to get you a config.log, the compile log, which includes where it attempts to run configure is in the paste.centos.org link above..

16:13 <karlp> here's config.log: https://paste.centos.org/view/98a01e9c

16:14 <karlp> am I correct in thinking that, right now, you can have openocd built against _either_ the internal libjaylink, if you don't have libjaylink explicitly selected in .config, or against libjaylink, if you did?

16:14 <karlp> are those currently the same version? so this "doesn't matter" ?

16:17 <PaulFertser> karlp: 0.12.0 will by default use external libjaylink, if it's not found, internal. Current master will by default (without special configure options) use only external libjaylink if available, else jlink support will be disabled.

16:17 <PaulFertser> Currently internal is the same 0.3.1 that's the last release from libjaylink upstream.

16:19 <PaulFertser> karlp: you pasted configure log but not from jimtcl which failed.

16:23 <karlp> https://paste.centos.org/view/cb5fd2b3

16:25 <karlp> ccache recursion gone wild somewhere...

16:25 <karlp> huh, 0.11 fails here in this tree as well.

16:26 <karlp> guess I haven't built it for a while :)

16:41 <PaulFertser> karlp: why is it ccache_cc ? On Jenkins I do export CC="ccache gcc"; export CXX="ccache g++"

16:41 <PaulFertser> Same for export CC="ccache clang"

16:43 <PaulFertser> And it looks like some other env variable sets ccache for you, and it gets appended.

16:48 <karlp> this is just "use ccache? y/n" in "Advanced config options" in menuconfig.

16:51 <PaulFertser> Interesting, never tried it.

16:53 <karlp> it's "default" for me, for years now.

16:55 <PaulFertser> I will try.

17:01 <PaulFertser> karlp: reproduced, will dig, thank you! Not sure if I need to put a note on github pull request now so that it's not merged by someone else?

17:09 <karlp> I've assigned them tomyself, so hopefully mangix et all will leave it for a couple of days at least.

17:09 <karlp> I don't use my powers much, but i have them, so I can work on these ones at least :)

17:10 <karlp> I will have "all day" tomorrow for it as well, but I'm not setup at home for this right now.

17:21 <PaulFertser> karlp: I have a workaround but I want to understand more, specifically, why it doesn't fail on jenkins.

17:24 <PaulFertser> OK, I see, it runs something like "ccache ccache x86_64-w64-mingw32-gcc ..." and that's ok for ccache...

18:03 <PaulFertser> karlp: https://github.com/ccache/ccache/blob/master/src/ccache.cpp#L2105 and https://github.com/ccache/ccache/blob/master/src/Util.cpp#L802 makes it treat anything that starts with "ccache" (including ccache_cc) to be a chain of ccache invocations which should be ignored and so the next argument is expected to be the compiler. I think OpenWrt is wrong here in calling the wrapper ccache_cc , as

18:03 <PaulFertser> that's breaking ccache assumptions.

18:54 <PaulFertser> I do not think jimtcl is in the wrong here trying to use ccache automatically if it's present in the PATH.

19:05 <PaulFertser> karlp: I propose to rename tools/ccache/files/ccache_cc to cc_cache (and same for cxx) and also make the corresponding change to rules.mk.

19:13 <PaulFertser> karlp: or even removing the wrappers altogether, why can't it be TARGET_CC:= "ccache $(TARGET_CC_NOCACHE)" ?

19:33 <karlp> take it up with openwrt. I'm not going to be the one solving that sory...

19:34 <karlp> I believe jim shouild use CC when it's been set, not go "hurhur, I are smrat, I'm going to do ccache $CC"

19:35 <karlp> but really, that's all an openwrt thing, I just don't feel comfortable merging it personally when it won't build like that :|

19:39 <Deneb> how does one access a global const from within a class?

19:41 <Deneb> I am getting 'fileTypes' is not defined'. 'fileTypes is defined as a global const within the same .js file.

19:42 <Deneb> oops, posten in the wrong channel! My apologies!

20:02 <karlp> PaulFertser: also, this is literally the first project I've had fail with ccache in openwrt... ever.

20:03 <PaulFertser> karlp: fair enough

20:03 <karlp> I packaged openocd-next for internal use at some stage, so I must have had it working v0.11.0-411-gc3993d318 at least.

20:04 <karlp> perhaps I didn't "clean" sufficiently when I was rebuilidng 0.11 earlier this afternoon.

20:05 <karlp> it's quite possible that openwrt's ccache wrappers are not as required as they were 10 years ago too... :)

20:05 <karlp> but yeah, not my circus there.

20:53 <PaulFertser> karlp: I'm going to send https://termbin.com/tvyr6 which allows to build OpenOCD with ccache enabled.

20:59 <PaulFertser> btw, libjaylink is not affected, so probably it can be merged before this ccache story is decided.

21:07 <karlp> true, I just wanted to test it running :)

21:13 <PaulFertser> This patch allows the build to happen nicely without any tweaks.

21:29 <karlp> it looks good to me, if tomorrow is clear I can re-reun my builds with your patch, and endorse it at least. but... :)