Re: debug-packages for aarch64?
On 2020/04/05 16:11, Jeremie Courreges-Anglas wrote: > On Sat, Apr 04 2020, Stuart Henderson wrote: > > I think it would be useful to have debug packages on aarch64. > > The main bulk build machines are powerful, and most of the machines > > users are running this on are not, so rebuilding things locally with > > symbols is quite a pain. > > Indeed rebuilding stuff on a rpi3 or a pinebook looks painful. exactly. > Seems safe. Another concern is the size increase on the mirrors, but > giving this a try is the easiest way to know for sure. ok jca@ fwiw amd64 debug-* are about 2.6G at the moment (down from 2.7G in February; since then more debug packages have been added but we also have dwz to shrink them). It will be interesting to see how much bigger the non-debug packages end up (due to the lack of handling for static libraries) as it will give a clue how bad that is on amd64 too.. I've discussed offlist with phessler, after the current bulk he's planning to give his a try with a local diff, and see how it goes before I commit anything.
Re: debug-packages for aarch64?
On Sat, Apr 04 2020, Stuart Henderson wrote:
> I think it would be useful to have debug packages on aarch64.
> The main bulk build machines are powerful, and most of the machines
> users are running this on are not, so rebuilding things locally with
> symbols is quite a pain.
Indeed rebuilding stuff on a rpi3 or a pinebook looks painful.
> They are working OK for me so far (N.B. I have only tested a handful)
> - I was sceptical before I tried (I wasn't sure it had enough of the
> gnu toolchain to handle them) but it's stripping and adding the debug
> links as expected.
>
> Any objections or OKs? (I'll review results from the next bulk and pull
> packages/backout before signing if it looks bad).
Seems safe. Another concern is the size increase on the mirrors, but
giving this a try is the easiest way to know for sure. ok jca@ fwiw
>
> Index: arch-defines.mk
> ===
> RCS file: /cvs/ports/infrastructure/mk/arch-defines.mk,v
> retrieving revision 1.71
> diff -u -p -r1.71 arch-defines.mk
> --- arch-defines.mk 4 Apr 2020 19:35:44 - 1.71
> +++ arch-defines.mk 4 Apr 2020 21:07:02 -
> @@ -40,7 +40,7 @@ GCC49_ARCHS = aarch64 alpha amd64 arm hp
>
> # arches where there is a C++11 compiler, either clang in base or ports-gcc
> CXX11_ARCHS = ${CLANG_ARCHS} ${GCC49_ARCHS}
> -DEBUGINFO_ARCHS = amd64
> +DEBUGINFO_ARCHS = aarch64 amd64
>
> .for PROP in ALL APM BE LE LP64 CLANG GCC4 GCC3 GCC49 MONO LLVM \
> CXX11 OCAML_NATIVE OCAML_NATIVE_DYNLINK GO \
>
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
Re: debug-packages for aarch64?
On Sat, Apr 04, 2020 at 10:29:36PM +0100, Stuart Henderson wrote:
> I think it would be useful to have debug packages on aarch64.
> The main bulk build machines are powerful, and most of the machines
> users are running this on are not, so rebuilding things locally with
> symbols is quite a pain.
> They are working OK for me so far (N.B. I have only tested a handful)
> - I was sceptical before I tried (I wasn't sure it had enough of the
> gnu toolchain to handle them) but it's stripping and adding the debug
> links as expected.
> Any objections or OKs? (I'll review results from the next bulk and pull
> packages/backout before signing if it looks bad).
Sounds very reasonable.
ok kmos
--Kurt
> Index: arch-defines.mk
> ===
> RCS file: /cvs/ports/infrastructure/mk/arch-defines.mk,v
> retrieving revision 1.71
> diff -u -p -r1.71 arch-defines.mk
> --- arch-defines.mk 4 Apr 2020 19:35:44 - 1.71
> +++ arch-defines.mk 4 Apr 2020 21:07:02 -
> @@ -40,7 +40,7 @@ GCC49_ARCHS = aarch64 alpha amd64 arm hp
>
> # arches where there is a C++11 compiler, either clang in base or ports-gcc
> CXX11_ARCHS = ${CLANG_ARCHS} ${GCC49_ARCHS}
> -DEBUGINFO_ARCHS = amd64
> +DEBUGINFO_ARCHS = aarch64 amd64
>
> .for PROP in ALL APM BE LE LP64 CLANG GCC4 GCC3 GCC49 MONO LLVM \
> CXX11 OCAML_NATIVE OCAML_NATIVE_DYNLINK GO \
>
debug-packages for aarch64?
I think it would be useful to have debug packages on aarch64.
The main bulk build machines are powerful, and most of the machines
users are running this on are not, so rebuilding things locally with
symbols is quite a pain.
They are working OK for me so far (N.B. I have only tested a handful)
- I was sceptical before I tried (I wasn't sure it had enough of the
gnu toolchain to handle them) but it's stripping and adding the debug
links as expected.
Any objections or OKs? (I'll review results from the next bulk and pull
packages/backout before signing if it looks bad).
Index: arch-defines.mk
===
RCS file: /cvs/ports/infrastructure/mk/arch-defines.mk,v
retrieving revision 1.71
diff -u -p -r1.71 arch-defines.mk
--- arch-defines.mk 4 Apr 2020 19:35:44 - 1.71
+++ arch-defines.mk 4 Apr 2020 21:07:02 -
@@ -40,7 +40,7 @@ GCC49_ARCHS = aarch64 alpha amd64 arm hp
# arches where there is a C++11 compiler, either clang in base or ports-gcc
CXX11_ARCHS = ${CLANG_ARCHS} ${GCC49_ARCHS}
-DEBUGINFO_ARCHS = amd64
+DEBUGINFO_ARCHS = aarch64 amd64
.for PROP in ALL APM BE LE LP64 CLANG GCC4 GCC3 GCC49 MONO LLVM \
CXX11 OCAML_NATIVE OCAML_NATIVE_DYNLINK GO \
