[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

RFC: fix non-strippability of -custom executables in a Debian-specific way



Hello,

I would like to apply to the ocaml package in Debian the patch I submitted at [1], even though it has been rejected upstream.

I still agree that -custom should be avoided when possible... but contrarily to Xavier, I don't think it is always possible *and relevant* in practice. I'm thinking of projects than provide single-use C stubs (i.e. stubs that are used only it its own executables). I've fixed ocsigen and coq in this regard... but I am upstream for those projects, and the changes were quite intrusive. It doesn't look serious to me to expect all our upstreams to go through all this trouble to support architectures where ocamlopt is not available, i.e. that only Debian seems to care about.

The non-strippability issue is still on-topic, as the recent #603580 [2] bug shows... and I wouldn't be surprised to see similar bugs in the future.

I still think that -custom should be forbidden for libraries by policy, but I'd like to relax the policy to allow -custom for executables that use their own "single-use" stubs.

As far as duplication of the interpreter goes: this is a more general issue with ocaml: native-code executables also embed the runtime, and all executables also embed the standard library. Anyway, this is a different issue that should be addressed by other means IMHO.

As far as compatibility goes: this patch might break the possibility to run the executable with a different runtime. I am dubious about the usefulness of this (maybe cross-compiling), even though it is technically possible now. And even with my patch, it will still be possible to compile seperately the runtime and the bytecode (using the -make-runtime/-use-runtime options of ocamlc), and do whatever combination one likes with them.

I am planning to upload ocaml 3.12.0 with this patch unless there is strong objection, but I would like to hear from other people's opinion (positive or negative).

[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=42;bug=256900
[2] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=603580


Cheers,

--
Stéphane


Reply to: