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

Re: [Pkg-mono-group] Bug#394418: question for ARM porters: incomplete arm v3 support in etch?



Hi, I am the maintainer of Mono and I want to add some (maybe useful)
information that may help into fixing this problem or at least
understanding it. (Warning: I am not a porter, and I do not know the
internals of Mono)

This build problem of Mono is a runtime problem that only occurs on
elara, netwinder, europa and smackdown.
During the build of Mono (it's a runtime which bootstraps it fails on
only those machines at runtime stage.

It first builds the JIT (Just In Time compiler) which translates the CIL
bytecode (*.exe or *.dll) into the ASM instructions.
Then the C# compiler (mcs) that is used during boostrap is running on
top of that runtime, so if the JIT emits instructions
that doesn't work correctly on that CPU it will fail in pretty random
ways, currently seen as SIGSEGV, SIGABRT or an FileNotFoundException
(which is bogus btw).
So what we see in the build logs of Mono, is that the application
mcs.exe which runs on top of the JIT crashes.

This issue is filed upstream under:
http://bugs.ximian.com/show_bug.cgi?id=78183

Here a report of builds that worked and builds that failed:

FAIL
----

elara: SIGSEGV
http://buildd.debian.org/fetch.cgi?pkg=mono&arch=arm&ver=1.1.18-3&stamp=1161343362

netwinder: SIGSEGV
http://buildd.debian.org/fetch.cgi?pkg=mono;ver=1.1.18-1;arch=arm;stamp=1161112787

netwinder: SIGABRT
http://buildd.debian.org/fetch.php?&pkg=mono&ver=1.1.17.1-4&arch=arm&stamp=1159074469&file=log&as=raw

europa: SIGABRT
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.17.1-2&arch=arm&stamp=1157817237&file=log

smackdown: SIGABRT
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.17.1-1&arch=arm&stamp=1157335424&file=log

FAIL of packages that use the Mono runtime (this shows its not a problem
of the compiler but runtime):
europa: Unhandled Exception: System.IO.FileNotFoundException: RSS.NET
http://buildd.debian.org/fetch.php?&pkg=blam&ver=1.8.2-2&arch=arm&stamp=1152341997&file=log&as=raw

elara: Unhandled Exception: System.IO.FileNotFoundException:
mscorlib.resources
http://buildd.debian.org/fetch.php?&pkg=ikvm&ver=0.28.0.0-1&arch=arm&stamp=1154063310&file=log&as=raw

europa: Unhandled Exception: System.IO.FileNotFoundException:
evolution-sharp
http://buildd.debian.org/fetch.php?&pkg=evolution-sharp&ver=0.11.1-1&arch=arm&stamp=1149043910&file=log&as=raw

SUCCESS
-------

toffee:
http://buildd.debian.org/fetch.cgi?pkg=mono;ver=1.1.18-3;arch=arm;stamp=1161939692

toffee:
http://buildd.debian.org/fetch.cgi?pkg=mono;ver=1.1.17.1-5;arch=arm;stamp=1159948410

cats:
http://buildd.debian.org/fetch.cgi?pkg=mono;ver=1.1.17.1-4;arch=arm;stamp=1159460374

cats:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.17.1-3&arch=arm&stamp=1158536080&file=log

smackdown:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.8-1&arch=arm&stamp=1152124452&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.6-4&arch=arm&stamp=1147240675&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.6-3&arch=arm&stamp=1147640592&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.6-2&arch=arm&stamp=1145751188&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.4-1&arch=arm&stamp=1142684613&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.2-1&arch=arm&stamp=1138914220&file=log

cats:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.13.1-1&arch=arm&stamp=1137330496&file=log

toffee:
http://buildd.debian.org/fetch.cgi?&pkg=mono&ver=1.1.12.1-1&arch=arm&stamp=1136143915&file=log

We can see here, that the same upstream version and same debian
revision, show different results between netwinder model
and cats model: it builds on cats and fails on netwinder (see 1.1.18-3
and 1.1.17.1-4).

As said I am not a porter, so I don't know the difference between cats
and netwinder, but AFAIK cats is v4l and netwinder is v3l.

Upstream tests and only has access to arm v5l and can't reproduce this
problem, as seen in the upstream bugreport.

Hope this helps.

-- 
Regards,

Mirco 'meebey' Bauer

PGP-Key:
http://keyserver.noreply.org/pks/lookup?op=get&search=0xEEF946C8

-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GIT d s-:+ a-- C++ UL++++$ P L++$>+++$ E- W+++$ N o? K- w++>! O---- M-
V? PS
PE+ Y- PGP++ t 5+ X++ R tv+ b+ DI? D+ G>++ e h! r->++ y?
------END GEEK CODE BLOCK------

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: