On Tue, Jun 16, 2020 at 11:24:11AM +0300, Adrian Bunk wrote: > On Mon, Jun 15, 2020 at 12:17:39AM +0200, Vctl Piotr Kolasinski wrote: > >... > > In my opinion the big problem may be w ith access to the real platform > > (currently I have access and possibilities but I don't know how long). > > Of course we have emulators like hercules (which I use from very long > > time) and know qemu s390 port (only with virtio, not tested by me), > > but it is probably not enough in power (as long as yo don't have very > > powerful emulation platform or use cross-compile). Anyway if Debian > > maintainers have access to valid build environment, I think you should > > not remove the architecture. > > Hardware is not the problem. > Forcing 1000 volunteers in Debian to support a port that has no porters > and no users is the problem. > > Debian has an s390x porterbox that is available to all Debian developers. > For normal package development this is sufficient. > > The s390x port has some unique problems. > And with a390x as release architecture package maintainers in Debian are > supposed to fix these problems in their packages if they want their > packages in the next Debian release. > > Forcing volunteers to do unpleasant work like porting to s390x is making > it a more attractive choice to stop contributing to Debian. I understand your point. I just started to read about Debian maintainers, developers, porteboxes etc., I will not discuss with you. My mail was sent because I use Debian in any possible place, advertise it, treat as stable, good alternative to other distros. From other site I didn't know about unique problems for that arch. > > s390x is the only big endian release architecture. > Big endian hardware has become exotic, and some of the younger > maintainers in Debian might have never seen big endian hardware. > Endian problems are common problems in packages, > and porting software to support big endian can be a real pain. > Is that not why Linux is so popular - portability?
> s390x is the only headless release architecture. > This was a real pain for the Debian GNOME maintainers already before > the last release, without any support from s390x porters on fixing > this issue.[1] I don't agree. Fact, that s390x doesn't have direct display doesn't mean that graphical tools are not used. Is really good way to thinking about Linux as desktop system only? In my daily work we have much more Linux boxes in VM (as host and target) then desktop system (where Windows is the king and never will change in next 20 years). I still remember that X Window architecture assume remote operations and local Xserver is only one of a few possible configurations. Of course if we assume, that we produce desktop system, the s390x is not composing with that. I use Debian in many instances - only two of them I access directly (using graphics), the rest is some kind of "server" (but I still can reach then graphically using vnc protocol). > > A port like s390x with unique problems is only sustainable when several > people with good knowledge of Debian, s390x hardware and the Linux > kernel have a long-term commitment of swiftly supporting everyone in > Debian with s390x problems. Good point - the question is why there is not so many people with "good knowledge of Debian" in Mainframe environment? How many of potential mainframe users know that Debian supports s390x architecture? If not many - why? About "s390x hardware and the Linux kernel" - I'm just only advanced user (I hope), but as I know - the kernel support in this area is mostly done by IBM. Does someone can tell me how many changes have to be done in Debian and what kind for the kernel? Does we really need many people in this area? > > IMHO it would be best if s390x would become a non-release architecture > in ports. My previous message was sent because I worry, that if this architecture will become not-release, after short time disappear (and some people abandon it or switch to Ubuntu - as long as it exists). > > Architectures in ports are autobuilt like release architectures, > but there is no pressure on the volunteers maintaining packages > in Debian to spend their time on supporting these architectures. > Other architectures like m68k, big endian powerpc, alpha, hppa > and ia64 that also tend to have one dedicated porter each but > not many users left are also in ports. Here is my lack of knowledge - I just started to dig Debian ecosystem as supporter, I don't know too much about preparing releases, building packages etc. Last time when I had problems with installation on Mainframe, I have learned many things about Debian installer, but ... I even don't know where to start reading about it be useful for others. Anyway I want to be active Debian user (partially on the Mainframe) and I want to help. Anyway thank for support till this time. Piotr
