Hi Helmut,
On 5 November 2024 at 06:28, Helmut Grohne wrote:
| On Mon, Nov 04, 2024 at 04:26:23PM -0600, Dirk Eddelbuettel wrote:
| > On 4 November 2024 at 22:18, Helmut Grohne wrote:
| > | Source: gsl
| > | Version: 2.8+dfsg-4
| > | Tags: patch
| > | User: debian-cr...@lists.debian.org
| > | Usertags: cross-satisfiability
| > |
| > | gsl cannot be cross built from source, because its dependencies on gcc
| > | and binutils cannot satisfied. They need "toolchain dependency cross
| > | translation" which has been unavailable until earlier this year.
| >
| > I read this paragraph three times and still do not understand. The
| > constraint was there for years, possibly a decade, and always satisfied.
|
| I vaguely feared that and tried to cut down unnecessary detail, but
| evidently went too far.
==:-)
| > What does "toolchain dependency cross translation" mean? What does it
| > change?
|
| If you depend on a compiler (including binutils), you currently are not
| precise what architecture you are compiling for. This used to be fine
| when we did native builds only as the implied architecture simply was
| the native one. Now when cross building, you mostly want to compile
| for the machine you are cross compiling for and sometimes want to
| compile for the machine you are building on (in order to run something
| compiled during build). This is an extra bit of information that was not
| needed earlier. Therefore, we will have to touch each and every
| dependency on a compiler in the archive and change it expressing this
| extra information. Roughly speaking the way forward is to append
| "-for-host" to the compiler package name for the cross compiler and
| "-for-build" when you want to compile a build tool to be run during
| build.
|
| Let me know if you want more detail and I'll expand.
No that's crystal clear. (And I have seen cross-compile used in earnest in a
few places. For R, the CRAN archive uses it a lot for the Windows UCRT
build. The arm64 builds for macOS can also be done this way etc.)
|
| > (I did see posts from you on debian-science but I haven't followed super
| > closely ...)
|
| That's quite related. It's a deep topic and d-science just happens to be
| affected early, but it really affects more of Debian, which is why I
| originally posted it to d-cross@l.d.o and d-toolchain@l.d.o only. I'm
| glad for feedback, but I also understand if you want to skip that part.
I will keep an eye on my debian/control files.
| > Thanks, I think I will fold it in. We can do with with the versioned
depends.
|
| I hope this is a typo and you meant s/with/without/.
Indeed. My bad. We have a custom 'fortunes' package with the R universe and
it had for years this:
> fortunes::fortune("Dirk can type")
I think we can reject the null hypothesis of "Dirk can type" at all
convential significance levels.
-- Dirk Eddelbuettel (after several users including himself had misspelled
Ubuntu as Umbutu or Ubunto)
R-help (April 2005)
>
Cheers, Dirk
--
dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
