gpsd-dev
[Top][All Lists]
Advanced

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

Re: [gpsd-dev] Bernd's CheckFunc Fix


From: Bernd Zeimetz
Subject: Re: [gpsd-dev] Bernd's CheckFunc Fix
Date: Wed, 04 Jan 2017 07:30:56 +0100
User-agent: K-9 Mail for Android

Hi,

short question - which scons version are you using?

Thanks,
Bernd

Am 4. Jänner 2017 02:15:46 MEZ schrieb Fred Wright <address@hidden>:

On Mon, 2 Jan 2017, Bernd Zeimetz wrote:

sorry, I missed that mail - please CC me if you need my attention or
poke me on irc.

Well, last time I tried CCing you, it made things worse. :-) The GPSD
listserver avoids forwarding to anyone explicitly listed as a recipient,
so CCing means that you *only* get the direct copy, and unless you've
whitelisted me in your spam filter, you don't get that one, either.

Trying a BCC this time, which may or may not help.

On 12/20/2016 12:35 AM, Fred Wright wrote:
[..]
4) Check out fde48f55 or later.

That change does not help, as we are dealing with compiler default
options here, not with those added in SConstruct.

More than just that - see below.

Of course it would still be nice to know how a warning became an error.

afaik default since 6.2.

Not in 6.3, or even in your own example.

% gcc -Q --help=warnings | grep enabled
-Waggressive-loop-optimizations [enabled]
[...]

Actually, -Wstrict-prototypes doesn't appear in this list.

Also, that says nothing about how warnings became errors. Warnings per se
don't cause configure checks to fail.

On Mon, 2 Jan 2017, Bernd Zeimetz wrote:

if you revert this commit, please ensure gpsd builds in Debian unstable
with the following options

export LDFLAGS="-Wl,-z,relro"
export CFLAGS="-g -O2 -fstack-protector-strong -Wformat
-Werror=format-security"
export CXXFLAGS="-g -O2 -fstack-protector-strong -Wformat
-Werror=format-security"
export CPPFLAGS="-Wdate-time -D_FORTIFY_SOURCE=2"

scons prefix=/usr systemd=yes strip=no dbus_export=yes
docdir=/usr/share/doc/gpsd libdir=/usr/lib gpsd_user=gpsd
gpsd_group=dialout debug=no


Testing "on two linux versions" does not help. Just reverting my commits

Indeed, which is why I asked for help testing.

You should be able to make it fail with gcc 6.2 on most platforms.

Nope (if you mean "at least 6.2"). I don't have a Debian unstable
install, but I did just add a Fedora 25 (with gcc 6.3) to my VM
collection. As long as Qt is turned off, it builds and passes all the
regression tests with no warnings. Though getting the options to the
almost-default state (i.e., except qt) requires installing a few more
packages.

I also tried it (after a full clean) with the above flag settings and
options, and it *still* built and tested cleanly. So I'm still unable to
reproduce the problem. And I looked specifically at some CheckFunc cases
in config.log (both successful and properly unsuccessful), which looked
like they behaved as intended.

So please revert the revert, or fix the issue on your preferred way.

In order to have a "preferred way", we need to understand what the real
problem is. :-)

Fred Wright

--
Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.
reply via email to

[Prev in Thread] Current Thread [Next in Thread]