V4 for musl#1012
Conversation
Needed for build against musl libc Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Needed for build against musl libc Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
fix warnings patches sent upstream Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
fix warnings patches sent upstream Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
fix growisofs.c:661:17: warning: implicit declaration of function 'major' growisofs.c:662:17: warning: implicit declaration of function 'minor' Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
fix
error: #warning redirecting incorrect #include <sys/poll.h> to <poll.h>
[-Werror=cpp]
1 | #warning redirecting incorrect #include <sys/poll.h> to <poll.h>
| ^~~~~~~
cc1: all warnings being treated as errors
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
fix ../../git/src/webrequest.cpp:15:10: fatal error: sys/unistd.h: No such file patch sent upstream Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
see patch headers for the list of fixes note that the recipe exhibits severe warnings patches sent upstream Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
v1 Try to make the code more portable. Fixes are listed in the patch header. Note: malloc/mallinfo is not available in musl (deprecated) patch sent upstream Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
ofgwrite has been updatet |
Patches accepted upstream This reverts commit c290b24.
|
Did you check this PR or you just merged it? Now you're patching your own enigma2 and tuxtxt instead of merging the related PRs: |
|
Hmm, did I miss something? I'm busy elsewhere for a few days, and have over 200 github notifications pending in my inbox... :( |
|
I think I've merged what I can merge, so I now need to revert these three commits? |
|
Yep you need to revert those commits and possibly more in future after eriksl/streamproxy#16 got merged. This PR should have been a draft not a real one and I told in prev ones to go for upstreams not provide patches. |
|
Ok, clear. |
|
Done. |
|
This commit should be reverted and applied as a patch against the repo: OpenPLi/enigma2@ad3d13e It's only the patch file. |
|
Also 147a8ff was a correct commit, not sure why @kueken tagged it as one to revert as it's the reason this whole thing exist. Some of the cleanups in d9fe777#diff-b3965358f2cc5704b05298e6d2989758e27610a5d57bdaaff5b6b0848e75f1d5 were also fine. What can I say, it's a mess ... |
|
Which is why I asked specificly if those three should be reverted, and you said "yes"? |
|
I was referring to my post and the related patches/commits: #1012 (comment) Maybe I got you wrong or you got me wrong, but I sent the correction for enigma2 repo 👍 |
|
I am sorry, these are fallouts of this PR. The problem here is IMHO that the recipes are not pinned to a git revision so you do not have control over mainstream changes: recipes are automatically updated to latest. There wasn't time to send the specific PR for OpenPLi, just the first for ofgwrite. I sent the patches upstream and the mainainers did just react too fast :) Good so! My 2 cents. |
|
Any package than needs specific pinning in the develop branch needs its SRC_REV listed in reporefs.conf, we don't pin in individual recipes. |
|
musl build now fails with Well, how to say, once this is done there will be a next problem for musl, especially for enigma2. :: musl has moved to 64bit time_t and does not play games like glibc The issue will impact four recipes: directfb_1.7.7, v4l-utils_1.16.5, busybox_1.31, enigma2 Ah, some good news then, I have backported a couple of classes and the whole musl libc + gcompat. For the moment I am battling with the serial cable, hoping it's not the connector on the vuduo2. Cheers |
|
Is the effort really worth? |
* Revert "enigma2: fx build with musl (#3014)" It's a patch file and should be applied first. This reverts commit ad3d13e. * enigma2: fix build with musl fixes: lib/base/thread.cpp:59:5: error: 'struct sched_param' has no member named '__sched_priority'; lib/base/e2avahi.cpp:154:37: error: missing sentinel in function call [-Werror=format=] lib/dvb/epgcache.cpp:400:2: error: 'PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP' was not declared in this scope; lib/dvb/epgtransponderdatareader.cpp:10:65: error: 'PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP' was not declared in this scope; lib/gdi/gpixmap.cpp:9:2: error: #error "no BYTE_ORDER defined!" lib/gdi/font.cpp:35:31: error: 'PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP' was not declared in this scope; lib/gdi/pixmapcache.h:11:9: error: 'uint' does not name a type; did you mean 'rint'? lib/gdi/pixmapcache.cpp:134:30: error: 'MaximumSize' was not declared in this scope lib/gdi/font.cpp:35:31: error: 'PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP' was not declared in this scope; lib/gdi/pixmapcache.h:11:9: error: 'uint' does not name a type; did you mean 'rint'? lib/gdi/pixmapcache.cpp:8:1: error: 'uint' does not name a type; did y lib/gdi/pixmapcache.cpp:134:30: error: 'MaximumSize' was not declared Plugins/Extensions/SocketMMI/src/socket_mmi.cpp:1: ...../usr/include/sys/poll.h:1:2: error: #warning redirecting incorrect include <sys/poll.h> to <poll.h> [-Werror=cpp] lib/gdi/font.cpp:35:31: error: 'PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP' was not declared in this scope; main/bsod.cpp:6:10: fatal error: execinfo.h: No such file or directory main/enigma.cpp:357:18: error: variable 'dump_malloc_stats()::mallinfo mi' has initializer but incomplete type main/enigma.cpp:357:32: error: invalid use of incomplete type 'struct dump_malloc_stats()::mallinfo' Signed-off-by: Andrea Adami <andrea.adami@gmail.com> Signed-off-by: persianpros <persianpros@yahoo.com> Co-authored-by: Andrea Adami <andrea.adami@gmail.com>
|
I wonder exactly the same. Also because the space you might save is relatively insignificant, we'll have to decide to drop boxes with 128Mb flash sooner rather than later... |
|
If you can save them why not? Let him go for it and don't lose hope just now. |
|
I don't want to stop him. But there are several things which work different: Maybe better start with some applications which are known work with musl. |
|
Me neither. But we have to be careful not to end up with a lot of complexity and instability, for not a lot of gain... |
|
So, about musl, I am nor the author nor a lobbyst :) As for OpenEmbedded, the outstanding work of Khem provides full musl-compatibility for oe-core and I'd say for meta-oe as well. It's maybe interesting to see how it helps the older devices like vusolo, having less (glibc-)binary drivers in the image. devtmpfs: mounted So before testing gcompat for dlopen() I must get to console :/ A.A. |
|
Looks like drivers don’t get loaded. |
|
About ubi, yes, something fails to open_volume() once mtd0 is already attached (initscripts or udev rule I think). Creating 4 MTD partitions on "brcmnand.0": |
5 participants
Apart dvd+rw-tools the patches have been cleaned and sent upstream.
See patch header for the fixed errors and commit log for the changes.
Thanks for reviewing it!
A.A.