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

Re: Help with mariadb-10.3 cross-building



user debian-cross@lists.debian.org
usertags 707750 + ftcbfs
affects 707750 + src:mariadb-10.3
thanks

Hi Otto,

On Thu, Jan 09, 2020 at 09:25:07PM +0200, Otto Kekäläinen wrote:
> Despite previous help from Helmut and some effort into cross-building,
> mariadb-10.3 does not currently seem to cross build due to a perl
> dependency.

Thank you for looking into this!

> > Architectures: arm64, armel, armhf, ppc64el, s390x
> > Summary: conflict between perl-base:$host and perl-base:$build
> 
> http://crossqa.debian.net/src/mariadb-10.3
> https://bootstrap.debian.net/cross_all/mariadb-10.3.html

When you look into satisfiability issues such as this one, the
bootstrap.d.n link is your primary resource. It tells you that perl is
the only dependency issue left for mariadb-10.3.

> https://qa.debian.org/dose/debcheck/cross_unstable_main_amd64/latest/packages/mariadb-10.3.html
> 
> If somebody wants to dig deeper into this, I would be happy to
> facilitate in patches and upload new versions. I also promise to buy a
> beer at FOSDEM to anybody who contributes to MariadB in Debian :)

Well, the perl dependency very likely just needs an ":any" annotation.
If mariadb-10.3 also builds perl bindings somewhere (which doesn't seem
to be the case from a first glance), you need to depend on "perl-xs-dev"
these days. You can just go ahead and add ":any" to your perl build
dependency.

However that's not getting you very far. Once you do that, you can
attempt to cross build mariadb-10.3, but it won't finish installing its
Build-Depends. Instead, it'll abort with an unpack error from dpkg while
attempting to unpack libreadline-gplv2-dev. The relevant problem is
already reported: #707750. If I remember correctly, mariadb-10.3 really
needs readline for both architectures (at least that's declared in
Build-Depends). So before moving on, that bug must be fixed. That bug is
also the reason why I went so silent on mariadb-10.3. It might have
looked like disinterest, but that isn't the case.

Someone should do a QA upload of readline5 fixing that bug. I don't
think I can promise to be the one doing so at this time.

Once this is fixed, I'll look into mariadb-10.3 again to figure out
whether there is more low hanging fruit.

Helmut


Reply to: