Hello,
27.12.2017 15:16, Boyuan Yang пишет:
> Package telegram-desktop FTCBFS on all architecture according to buildd
> logs[1].
>
> There are all kinds of reasons about build failures across different
> architectures, including cc1plus internal compiler error, timeout, vmem
> exhaustion, etc.
>
> Please investigate into this issue. I'm not sure what we can do, perhaps
> the best plan is to force non-parallel compilation (instead of make -j4
> or whatsoever on buildds).
I already found a fix[1] for memory exhaustion by GCC. If we look at a
file where the RPL_CONSUMER_TYPE_ERASED_ALWAYS macro is used[2], we
discover that the replacement of the auto keyword with more specific
type solves the problem.
But unfortunately there is another issue with linking on all
architectures except for amd64 and i386. Once a solution is found, I'll
prepare a new version of the package.
That's example of the error:
obj.target/liblinux_glibc_wraps.a(linux_glibc_wraps_64.o): In
function `__wrap_clock_gettime':
./obj-powerpc64le-linux-gnu/./Telegram/SourceFiles/platform/linux/linux_glibc_wraps_64.c:27:
undefined reference to `clock_gettime@GLIBC_2.2.5'
/usr/bin/ld: Telegram: No symbol version section for versioned
symbol `clock_gettime@GLIBC_2.2.5'
/usr/bin/ld: final link failed: Nonrepresentable section on output
collect2: error: ld returned 1 exit status
I don't think that non-parallel build let us avoid the errors.
[1]:
https://anonscm.debian.org/git/collab-maint/telegram-desktop.git/commit/?id=4dc4aadc8497a996f91d75fa7d8b64884cf8b54c
[2]:
https://anonscm.debian.org/git/collab-maint/telegram-desktop.git/tree/Telegram/SourceFiles/rpl/consumer.h?id=4dc4aadc8497a996f91d75fa7d8b64884cf8b54c#n629