GNU bug report logs -
#45633
bcc and bpftrace require kernel headers from system
Previous Next
To reply to this bug, email your comments to 45633 AT debbugs.gnu.org.
Toggle the display of automated, internal messages from the tracker.
Report forwarded
to
bug-guix <at> gnu.org
:
bug#45633
; Package
guix
.
(Sun, 03 Jan 2021 18:33:03 GMT)
Full text and
rfc822 format available.
Acknowledgement sent
to
John Soo <jsoo1 <at> asu.edu>
:
New bug report received and forwarded. Copy sent to
bug-guix <at> gnu.org
.
(Sun, 03 Jan 2021 18:33:03 GMT)
Full text and
rfc822 format available.
Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):
Hi Guix,
When I added bcc and bpftrace to guix I didn't think much of which
version of the kernel headers would be used. After a few kernel
updates, it has become clear that the current-system kernel headers need
to match the kernel headers the packages are compiled with. Can they be
provided with a variant property? The ocaml packages that need
different ocaml versions seem like a nice model on how to do the kernel
headers.
Thoughts?
- John
Information forwarded
to
bug-guix <at> gnu.org
:
bug#45633
; Package
guix
.
(Sun, 03 Jan 2021 21:41:02 GMT)
Full text and
rfc822 format available.
Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
John,
This reminds me of <http://issues.guix.gnu.org/45592>.
John Soo 写道:
> Hi Guix,
>
> When I added bcc and bpftrace to guix I didn't think much of
> which
> version of the kernel headers would be used. After a few kernel
> updates, it has become clear that the current-system kernel
> headers need
This should probably be booted-system.
> to match the kernel headers the packages are compiled with. Can
> they be
> provided with a variant property? The ocaml packages that need
> different ocaml versions seem like a nice model on how to do the
> kernel
> headers.
I know exact kernel headers are needed at toolchain run time, when
building eBPF programmes. Are they really needed at toolchain
build time, too? That sounds wrong (but there are more things
fundamentally wrong with eBPF on Linux; ask me to rant about
CONFIG_IKHEADERS). Still, could pointing bcc/bpftrace to
booted-system headers at run time not suffice?
Kind regards,
T G-R
[signature.asc (application/pgp-signature, inline)]
Information forwarded
to
bug-guix <at> gnu.org
:
bug#45633
; Package
guix
.
(Sun, 03 Jan 2021 21:41:02 GMT)
Full text and
rfc822 format available.
Information forwarded
to
bug-guix <at> gnu.org
:
bug#45633
; Package
guix
.
(Mon, 04 Jan 2021 18:11:02 GMT)
Full text and
rfc822 format available.
Message #14 received at 45633 <at> debbugs.gnu.org (full text, mbox):
Hi Tobias,
Tobias Geerinckx-Rice <me <at> tobias.gr> writes:
> John,
>
> This reminds me of <http://issues.guix.gnu.org/45592>.
Hm, maybe. How does it remind you of zfs? It seems like zfs needs the
libraries from util-linux, right? bcc and bpftrace need the kernel
headers. The following should illustrate there is nothing to load
dynamically, right?
find $(guix build linux-libre-headers) -name '*.\.so*'
> This should probably be booted-system.
Yes, definitely, thanks. s/current-system/booted-system/
> I know exact kernel headers are needed at toolchain run time, when
> building eBPF programmes. Are they really needed at toolchain
> build time, too?
I am not sure what the headers are used for but they are indeed the .h
files (tree $(guix build linux-libre-headers)). I would assume they are
used by the gcc-toolchain when compiling the package as #includes.
What do you think?
- John
This bug report was last modified 4 years and 27 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.