Hi,

Upstream seems to be using both a testing m4 script grabbed from
Google[0] **and** a hardcoded list to determine whether
-fstack-protector should be enabled. It's strange that the test was
passed on hppa[1] and alpha[2]. Maybe the testing case is broken?

I may dig into it afterwards, but I can't find any alpha/hppa machine :-(

+1 for using `--disable-ssp' globally and depend on dpkg-buildflags,
if it works. Yet the problem may need be solved upstream later.

--
Regards,
Boyuan

[0] 
https://github.com/shadowsocks/shadowsocks-libev/blob/master/m4/stack-protector.m4
[1] 
https://buildd.debian.org/status/fetch.php?pkg=shadowsocks-libev&arch=hppa&ver=2.4.7%2B20160630%2Bds-1&stamp=1467501230
[2] 
https://buildd.debian.org/status/fetch.php?pkg=shadowsocks-libev&arch=alpha&ver=2.4.7%2B20160630%2Bds-1&stamp=1467501341
--
杨博远 / Boyuan Yang <[email protected]>
Homepage: https://hosiet.me/
Twitter: @hosiet


2016-07-04 4:56 GMT+08:00 Aaron M. Ucko <[email protected]>:
> Source: shadowsocks-libev
> Version: 2.4.7+20160630+ds-1
> Severity: important
> Justification: fails to build from source
>
> Builds of shadowsocks-libev on alpha and hppa have been failing:
>
>   utils.c:1:0: error: -fstack-protector not supported for this target 
> [-Werror]
>
> It looks like this instance of -fstack-protector comes from upstream's
> build system, which doesn't account for these architectures' lack of
> support for it.  Moreover, on the remaining architectures, dpkg-buildflags
> already supplies -fstack-protector-strong, so -fstack-protector either is
> redundant or results in less strong protection.
>
> Ideally, upstream's build system would test that -fstack-protector
> actually works, rather than hardcoding a list of architectures on which
> it's expected to.  (It would be even better if the build system got out of
> the way if it noticed -fstack-protector or -fstack-protector-strong
> already in CFLAGS.)
>
> As it is, I would suggest unconditionally configuring with --disable-ssp
> and deferring to dpkg-buildflags.
>
> Could you please take a look?
>
> Thanks!
>

Reply via email to