* [gentoo-musl] [GSoC] native-clang: daily report 07/01
@ 2016-07-01 15:42 Lei Zhang
2016-07-04 9:03 ` [gentoo-musl] " Lei Zhang
0 siblings, 1 reply; 3+ messages in thread
From: Lei Zhang @ 2016-07-01 15:42 UTC (permalink / raw
To: Luca Barbato, soc-admins, gentoo-musl
Plan for today:
- write ebuild for NetBSD's crtbegin/end
- write ebuild for libc++abi
Progress:
- write ebuild for crtbegin/end
libc++abi's situation is a bit complicated. To configure it, you need
the specify paths to libc++ and LLVM's source. libc++ has a similar
issue: it also needs to know the path to LLVM's source. Actually
libc++'s ebuild maintainer wrote a Makefile to circumvent its cmake
system, which works without knowing LLVM's source path. I guess libc++
doesn't really need LLVM's source; it's just some quirk of their cmake
system.
But I can't tell yet if libc++abi really needs libc++ or LLVM's
source. If not, perhaps I can write a Makefile just like libc++'s;
otherwise, maybe I need to keep LLVM's source somewhere after
installing it.
Although libc++, libc++abi, etc support out-of-tree build, it looks to
me they're not decoupled from LLVM enough as they're supposed to be.
Sigh...
Lei
^ permalink raw reply [flat|nested] 3+ messages in thread
* [gentoo-musl] Re: [GSoC] native-clang: daily report 07/01
2016-07-01 15:42 [gentoo-musl] [GSoC] native-clang: daily report 07/01 Lei Zhang
@ 2016-07-04 9:03 ` Lei Zhang
[not found] ` <0f0d2d59-c3b6-65e0-8ab5-5acf48d3d9f6@gentoo.org>
0 siblings, 1 reply; 3+ messages in thread
From: Lei Zhang @ 2016-07-04 9:03 UTC (permalink / raw
To: Luca Barbato, soc-admins, gentoo-musl
2016-07-01 23:42 GMT+08:00 Lei Zhang <zhanglei.april@gmail.com>:
> Plan for today:
> - write ebuild for NetBSD's crtbegin/end
> - write ebuild for libc++abi
>
> Progress:
> - write ebuild for crtbegin/end
>
> libc++abi's situation is a bit complicated. To configure it, you need
> the specify paths to libc++ and LLVM's source. libc++ has a similar
> issue: it also needs to know the path to LLVM's source. Actually
> libc++'s ebuild maintainer wrote a Makefile to circumvent its cmake
> system, which works without knowing LLVM's source path. I guess libc++
> doesn't really need LLVM's source; it's just some quirk of their cmake
> system.
>
> But I can't tell yet if libc++abi really needs libc++ or LLVM's
> source. If not, perhaps I can write a Makefile just like libc++'s;
> otherwise, maybe I need to keep LLVM's source somewhere after
> installing it.
It turns out libc++abi doesn't really need LLVM's source, but *does*
depend on libc++'s headers. I managed to write a Makefile for it to
circumvent cmake, providing libc++'s headers can be found somewhere.
The problem is: lib++ is supposed be built after libc++abi. My thought:
- if libc++ is already installed (e.g. against libcxxrt), installing
libc++abi has no impact on libc++
- otherwise, installing libc++abi also installs libc++, and libc++
will be configured against libc++abi
Or alternatively: replace libcxxrt with libc++abi in portage repo, and
always install libc++ and libc++abi together, just as clang and LLVM
do.
Which makes sense more?
Lei
^ permalink raw reply [flat|nested] 3+ messages in thread
* [gentoo-musl] Re: [GSoC] native-clang: daily report 07/01
[not found] ` <0f0d2d59-c3b6-65e0-8ab5-5acf48d3d9f6@gentoo.org>
@ 2016-07-04 12:27 ` Lei Zhang
0 siblings, 0 replies; 3+ messages in thread
From: Lei Zhang @ 2016-07-04 12:27 UTC (permalink / raw
To: Luca Barbato; +Cc: soc-admins, gentoo-musl
2016-07-04 17:55 GMT+08:00 Luca Barbato <lu_zero@gentoo.org>:
> On 04/07/16 11:03, Lei Zhang wrote:
>> Or alternatively: replace libcxxrt with libc++abi in portage repo, and
>> always install libc++ and libc++abi together, just as clang and LLVM
>> do.
>>
>> Which makes sense more?
>
> The libc++ + libc++abi is less nice but probably more straightforward.
Actually the current ebuild for libc++ supports both libcxxrt and
libsupc++ (part of libstdc++) as the ABI library. Which one to use is
controlled by USE flag "libcxxrt". If libc++ is already installed
without this flag, installing libcxxrt afterwards will have no impact,
which is similar to my first proposal.
Perhaps we could add another USE flag "libc++abi"; which one of the
three ABI libraries to use is then controlled by users.
Lei
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2016-07-04 12:27 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-07-01 15:42 [gentoo-musl] [GSoC] native-clang: daily report 07/01 Lei Zhang
2016-07-04 9:03 ` [gentoo-musl] " Lei Zhang
[not found] ` <0f0d2d59-c3b6-65e0-8ab5-5acf48d3d9f6@gentoo.org>
2016-07-04 12:27 ` Lei Zhang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox