1234774 - [Aries][Build] Failed to build on Mac OS X, elf.h file not found

Status: RESOLVED WONTFIX

(Firefox OS Graveyard :: GonkIntegration, defect)

Description

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

Same as Bug 1027682, I think this affects all devices.

Comment 1

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

Attachment: fix-elf-header

cherry-pick from https://github.com/mozilla-b2g/codeaurora_kernel_msm/commit/48b6a2c9fde18f2f7d9abddf05261c498c2bb2ba

Comment 2

[Michael Wu [:mwu]]

Comment on attachment 8701377 [details] [review]
fix-elf-header

Shinano review -> Alex

Comment 3

[:gerard-majax]

That's for KK right? I had a side project of getting that to build on Mac, but I have not had time to work on that for quite some time now :(

Comment 4

[:gerard-majax]

Wei Cheng, would you mine check also the status of L build regarding that? I don't have a mac, the only one I have is at the office and obviously with christmas coming I won't be near that soon :)

Comment 5

[:gerard-majax]

Comment on attachment 8701377 [details] [review]
fix-elf-header

looks good, thanks

Comment 6

[:gerard-majax]

Right, after thinking about that, I also remember that several people reported the build to complete properly on OSX but the resulting boot image was not booting. As far as I know, the kernel itself is fine. Do you have something booting after that?

Comment 7

[Thomas Nguyen (:tnguyen)]

I think elf.h should be identical with external/elfutils/libelf/elf.h. Correct me if I am wrong?
If yes, could we just add a script to copy external/elfutils/libelf/elf.h to /usr/local/include/ ?(it will work for both L and KK)

Comment 8

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

(In reply to Alexandre LISSY :gerard-majax from comment #4)
> Wei Cheng, would you mine check also the status of L build regarding that? I
> don't have a mac, the only one I have is at the office and obviously with
> christmas coming I won't be near that soon :)

Ok, I'll try this on shinano-l too.

(In reply to Alexandre LISSY :gerard-majax from comment #6)
> Right, after thinking about that, I also remember that several people
> reported the build to complete properly on OSX but the resulting boot image
> was not booting. As far as I know, the kernel itself is fine. Do you have
> something booting after that?

I never dive into it yet, but I did notice that even if there are some errors occurred during building boot image, the building process still regards it as succeed.
(e.g.: (BSD-style)gnutar does not support --owner=0, stat options ... etc.)
Maybe this is the one of the reasons.

(In reply to Thomas Nguyen[:tnguyen][:thomas][:nguyen] from comment #7)
> I think elf.h should be identical with external/elfutils/libelf/elf.h.
> Correct me if I am wrong?
> If yes, could we just add a script to copy external/elfutils/libelf/elf.h to
> /usr/local/include/ ?(it will work for both L and KK)

I think this is slightly strange, that a "build" process makes a side effect to global file system.

Comment 9

[Thomas Nguyen (:tnguyen)]

(In reply to Thomas Nguyen[:tnguyen][:thomas][:nguyen] from comment #7)
> I think elf.h should be identical with external/elfutils/libelf/elf.h.
> Correct me if I am wrong?
See there're some differences of your import elf.h and external/elfutils/libelf/elf.h. Does it have any effect?
Thanks

> I think this is slightly strange, that a "build" process makes a side effect
> to global file system.
Agree. Thanks

Comment 10

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

Attachment: aries-l fail log

aries-l build failed before I can test this patch.

Comment 11

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

(In reply to Thomas Nguyen[:tnguyen][:thomas][:nguyen] from comment #9)
> See there're some differences of your import elf.h and
> external/elfutils/libelf/elf.h. Does it have any effect?
> Thanks

Their content are different, but both work for me.
I can't tell what the difference.

Comment 12

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

(In reply to Wei-Cheng Pan [:wcp] [:wcpan] [:legnaleurc] from comment #10)
> aries-l build failed before I can test this patch.

Looks like the version check in build/core/combo/HOST_darwin-x86.mk needs to be update as well, because I'm using 10.11.2
But I still hit bug 1194687, can't find the header.

Comment 13

[:gerard-majax]

If AOSP does not support that version, should we care about that asap ? Maybe not.

Comment 14

[:gerard-majax]

Finally got time to play again with that on the Mac that sits on my desktop. It's 10.10 though, but we will see. I am repo syncing for now :)

Comment 15

[:gerard-majax]

(In reply to Wei-Cheng Pan [:wcp] [:wcpan] [:legnaleurc] from comment #12)
> (In reply to Wei-Cheng Pan [:wcp] [:wcpan] [:legnaleurc] from comment #10)
> > aries-l build failed before I can test this patch.
> 
> Looks like the version check in build/core/combo/HOST_darwin-x86.mk needs to
> be update as well, because I'm using 10.11.2

I also hit that issue on 10.10 with KK port.

Comment 16

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

(In reply to Alexandre LISSY :gerard-majax from comment #15)
> I also hit that issue on 10.10 with KK port.

For KK port I just ust SDK 10.9 (MAC_SDK_VERSION='10.9' ./build.sh) and it works.
Not sure if we should keep chasing the SDK version or not.

Comment 17

[:gerard-majax]

Looks like AOSP supported kernel by Sony already solved it more nicely? https://github.com/mozilla-b2g/codeaurora_kernel_msm/blob/sony-aosp-l/Android.mk#L125

Comment 18

[:gerard-majax]

Wei-Cheng, would you mind verifying comment 17? Maybe it's incomplete/we are missing some of the repos, but that line really make me thinking it should work on L. And we could do something similar on KK which would feel less hacky

Comment 19

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

Attachment: 0001-add-compile-flag.patch

I can confirm this works well.

The attached patch is a draft, I'm wondering if there is a better way to put these platform dependent flags into one file, like build/core/combo.

Comment 20

[:gerard-majax]

Wei-Cheng, thanks! Can you confirm this is the only patch needed on a clean Z3c KK port tree to get a buildable system and booting kernel ?

Comment 21

[Wei-Cheng Pan [:wcpan] [:wcp] [:legnaleurc] (left Mozilla)]

We still need the patch from bug 1234986 (either Thomas's patch or mine) to build a booting kernel.

Comment 22

[BMO Automation]

Firefox OS is not being worked on