SGI: Development

Yes to everything you guys have said but my point was that things like gtk and qt use language features that MIPSPro does not have, if we could add those features by patching MIPSPro we could compile latest versions of those widget sets, and all the modern killer apps like chrome and chromium and firefox. But since we're never likely to get our hands on the MIPSPro source code the only way we can do that is to patch the MIPSPro binary. Hence my statement that doing so would be either difficult or impossible. It's always been a mystery to me why no sgi engineers have ever outed themselves on here. Gotta be at least a few of them who are still fans of this stuff.

dexter1 wrote: dissing opensource projects is not the way to increase our enjoyment and productivity on our SGI's. So let's not do that, but instead focus on what does work

that goes hand in hand. to know what works fine we need to know what sucks or vice versa

foetz wrote: that goes hand in hand. to know what works fine we need to know what sucks or vice versa

Yes, the point is not dissing for fun but dissing to forestall the "we need mooodern software !" remarks.

A lot of modern software is ghastly. I don't think we need that. Although if everybody else wants that, that's fine too.

Dexter, you haven't been building open source on Irix for a few years, I think you are in for a shock. Of course, you can just fix their stupidities so it won't be a major barrier for you. But .....

I don't want to sound like a wet blanket, especially when we get another code warrior returning to the fold, but seriously, if the users don't step up their game, this isn't going anywhere. I hope this is the first sign of a rebirth of interest in actually using SGI computers

About the browser problem, something to consider -- I don't have an Irix browser problem. I rdesktop across the room and steal cycles from the Assistant. That's not what I'd like to do best but you know what ? It works. Now I can concentrate on something more productive than the endless fight with the Mozilla jerks.

Another possibility would be Proxomitron. I am not sure which is more difficult, recreating a recoding proxy or fighting with fireflop and gtk2 and all the other ridiculous dependencies involved. It's not just fireflop - it's everything it sits on, too. That's a massive undertaking because the people involved no longer even pretend to care about anything but Windows, the latest Loonix ("just aptyumget the 30 newest packages !") or maybe OS X. Firefox 3.5 would be cool, but ... the other side of that is, by the time it works the web will probably be totally useless, with the few remaining decent websites working fine in Mosaic

How about a list of What Do People Want to See ? for starters ? Here's mine :

For strictly nekoware, testing all the basic dependencies and getting them into /current would be YAY ! Nekoware was always so reliable, you could put anything from /current into your box and know it was good. We've fallen off that horse

Again about nekoware, maybe re-examine and codify the underlying conventions ? For example, I would now put 'release notes' as a default install. They are very useful. But html docs, puhlease. and the 58,000 languages that fossies love to install, sheesh. Some people here still have Indigo2's with 8 gig drives. Or smaller. I don't think we need Tibetan or atk or a lot of the other Copy-Mickeysoft stuff.

Programs :

Newer MPlayer. The old one works super but (supposedly) the newer versions have more support for multiple processors. Indigo guys won't care but Octane people would want that. Axatax has already done a lot on this, getting it over that last hurdle would be great. It's one of my favorite programs, works better in Irix than any media player I've used on Windows.

A word processor. Framemaker is good but definitely overkill for a letter to mom. Ted is okay but some smoothing of the rough edges would be nice. I found Axene source, that might be a nifty project. We did some playing with Maxwell without final success. If it were Indigo Magicked that might be a cool useful program. Open Office is ghastly.

Axene also has a spreadsheet that can (supposedly) read Excel files. Not all that sexy but if you use the computer, being able to read spreadhseets is useful.

A newer and more stable Dia. I've used the older one, it works and it's useful but it's not really up to snuff.

A newer GIMP ? That's a popular program and SGI's are supposed to be about graphics and the gimp was always pretty okay before. Don't know if they've gone goofballs since but pretty sure a lot of people would use that.

I don't know if people do ftp anymore, but I messed with Axyftp a little, it wouldn't take too much to truly Irrixify it. It's already Motif, just some rework to use more Indigo Magic widgets would make it into a real Irix app. There are more older Motif utilities like that out there, too. Or programs like Pho. Pho is gtk2 but it is so simple, if it were jacked up and Motif slid underneath, it's very useful for paging through directories of photos.

A lot of people would like some games. Fair enough. The Transport Tycoon looked like a natural, until I tried to build it. Here I disagree with you, dex. Those people don't need to be dissed. They need to be taken out behind the woodshed and whipped with a leather belt. Go try to build that before you tell me you disagree. But games, for the casual user, that's a definite yes ?

You'd know more than I about this but to me, it seems that putting a lot of effort into Pango and Cairo and GTK2 is not .... productive ? That group of people has gone off the deep end, their code is no longer portable in any way, you could spend your life trying to keep up but what do you get in the end ? A crappier version of something that worked better before. You get 10% more glitz for 60% more cpu but we ain't got that much cpu to waste on nonsense. Concentrating on a few useful programs and really optimising them for Irix, like y'all did with MPlayer, would be a better approach ?

Although all that seems to be dead in the water now, the bros are infatuated with smartphones these days, so maybe those basics will stay still long enough to be worth working on ?

I think the GIMP is enmeshed with a pretty recent gtk2, but I would think that some of the folks around here with gcc 4.7 would be able to compile them both. Hmmm... Maybe they're in /beta but it's far too late at night for me to look, perhaps in the morning. If I haven't forgotten. Getting old sucks...

vishnu wrote: I think the GIMP is enmeshed with a pretty recent gtk2...

You're right. What was I thinking ? that they would concentrate on the graphics instead of changing the toolkit every week ? what an idiot

http://feenix.darktech.org/files/gtk2.mpg

hamei wrote: ...The support at user level sucks. You can't get the "I wuvs my sgi ! it's the kewlest machine ever and I wanted one ever since grade school, why don't those meanies at sgi Release the Source ?" group to even try out the most basic apps and give feedback. Developers can't do everything.

And why should they, if no one even bothers to try it and report back ? It's like pissing into the wind.

You know what? I agree with you - despite being one of the more recent hobbyists rather than a professional user. However, I believe that Nekochan needs both types. If you use SGIs purely for professional purposes, it's only natural that as they get older, more outdated and less supported, professionals will often turn to newer/supported machines. Not all of them, but certainly a sizeable number who used to rely on SGIs. Why spend weeks trying to bludgeon something into shape so it'll compile under IRIX, when you can just run it through GCC on a modern Linux machine and be up and running in 15 minutes? If you want newer software easily, IRIX isn't much good these days.

Generally, the only people left using SGI machines are people who love them. Either professional users that don't like the taste of modern Linux and who still love the simplicity and cleanliness of IRIX, or hobbyist users who have managed to buy/luck into an ecosystem that was financially prohibitive back in SGI's golden days.

hamei wrote: But we don't have a community. It's like pulling teeth to get one person to say "hey, the newest version of libjpeg works for me."

The big problem here is that it's very difficult to know whether some of the underlying libraries work unless you're a developer. I can take the latest build of Dillo, test it out and provide feedback because it's an actual app I can run, use and try out - and I have done this. But for an underlying library, they're difficult to test without knowing how much (or little) use of them an end-user package makes.

What I'd propose here is that the devs who manage to compile these try to take a bit more ownership of getting them into current. Rather than just popping them into Incoming, mentioning it quickly, then leaving it at that, push for the testing. Tell us some example ways we can confirm the functioning of these packages. For example...

"Hey, I've just compiled the latest version of libjpeg. It seems to work fine on my Fuel with IRIX 6.5.30, but I've not tested it on other platforms. GIMP makes good use of this library, so if people could open/save/manipulate some different sized JPEGs in GIMP on an Octane, Indigo2 etc. running earlier versions of IRIX and report back, that'd be great. Trying a few other packages that make use of libjpeg would of course also be useful".

As an end user, I now have some info on how to test this. I can try it out, experiment a bit, report back with useful feedback. If you want people who aren't developers to report back and to help with testing, you have to help them to help you. If you just say "newest zlib is in incoming" and leave it at that, getting feedback on that is always going to be a challenge.

hamei wrote: A lot of people would like some games. Fair enough. The Transport Tycoon looked like a natural, until I tried to build it.

Yeah. It seems that 0.5.3 compiles quite easily - although it does crash with a Bus Error when using the map. All other functions are fine though. For 0.6.0 and above, they changed the configuration system, and I cannot get it to complete. I spent a few hours trying to work through it yesterday, but when you have a 10,000 character input string to sed which is being rejected for "too many arguments", it's difficult to know where to begin with fixing it unless you really know what you're doing. As you say, annoying!

Overall, yes I'm admittedly one of hamei's enthusiast users. My PC is my main every-day machine, not my SGIs. However, I'm always looking for things that I can do on them instead of my PC. I've spent a chunk of time recently playing a couple of ScummVM games on them. I could do that on my PC, but I prefer to do it on my SGIs if I can. Similarly if I can get OpenTTD working, I'd love to spend a good few hours having fun with that under IRIX. I do like to use my SGIs for proper work as well. I rip vinyls on my Indigo2, enjoy odd bits of perl scripting via NEdit etc. Hence, older retro games, game engines and emulators are things I'd like to see, plus simple and useful productivity tools.

Anyway, that's my 2 cents!

Trippynet wrote: Overall, yes I'm admittedly one of hamei's enthusiast users.

I apologize if you thought I was being negative. Enthusiasts are great, too. I was meant to be just explaining where I was coming from. Of course other people have other viewpoints.

No offense taken, I fully understand your point. Enthusiasts are nice, but of rather limited use when it comes to porting/testing software

I'll try my hand at Maxwell again, as I recall I managed to get every object file to compile but the makefile did something really stupid with them that resulted in no useable binary. Yeah, I'm remembering it now, that makefile was the result of a hand-rolled Makefile.in that made no sense, I'll see if I can't doctor it up.
"Who are you?"
"I'm the Doctor!"
Said every episode of Doctor Who ever...

vishnu wrote: What we should do n order to compile this "modern" shite on our SGI's is to write and compile binary patches to MIPSPro that add the functionality that C and C++ now possess, that MIPSPro doesn't. Difficult? Yes. Impossible? Probably. Now, who's with me? Into the fray!

WELL IT'S ALREADY BEN DONE! Let me explain. Oh about three months back I was busing looking into getting QT 4.8.5 working so I could build QTWeb. At this time I found out the areas the MIPSPro7.4m and GCC4.7 compilers were weak in. The Mipspro compiler was very fast,stable but the didn't support the latest C++11 syntax. The GNU compiler on the other hand supported C++11 syntax but while attempting the QT code would generate "internal compiler errors" plus it was very slow...It took twice as long to compile large C++ programs compared to Mipspro. At this time I had a choice of looking for a different compiler or start debugging the GNU compiler. I stumbled across LLVM which support MIPS big and little endian. I currently works on Mips Linux so I needed to get it to work on SGI. In the process of configuring it the configuration test programs said I was running "CLANG" but with obsolete version 6.0 of stdc++ libraries ...What the heck!
I did some checking and found out the front end code of MipsPro was acquired by Apple. Apple enhanced it, changed the name to "CLANG" and hooked it up to LKVM as the backend code generator and posted all the changes back into LLVM site. The LLVM fully supports three targets ...mips,intel and arm.
A group on the internet posts builds of the Linux kernel for the MIPS le and be built using the LLVM compiler. The kernel runs is reported to run faster than one built with GCC.
What this means right now people are basically running an updated version of our MipsPro compiler under MIPS Linux!
I think we should get the compiler working back under SGI!

I think you mixed up a few things as far as MIPSpro goes, at least.

The MIPSpro frontend (source parser) was licensed from the Edison Design Group . EDG supplied various other compiler vendors at the time and I believe the Intel Linux compilers still use it.

The MIPSpro intermediate (WHIRL) and backend was written by SGI. When SGI moved to IA64 years later, they added an IA64 code generator, and yanked out all the obsolete MIPS bits.

They open-sourced various parts of the compiler, but couldn't release the EDG frontend because it wasn't theirs. So they welded a GNU frontend on it which generates WHIRL instead of RTL. This MIPSpro-for-IA64 was released as Pro64 and was later renamed Open64 .

Open64 served as a research vehicle for a couple of years, new targets for x86-64 and MIPS (not related to MIPSpro!) were added, but seems to be dead now.

As far as I'm aware, clang doesn't use the EDG frontend. It would certainly be interesting to get a MIPS BE target of clang up and running on IRIX, but the IRIX implementation of the ELF format has a couple of quirks which might make that a challenge.

That would be fantastic if it could be done. A modern/updated MIPSPro compiler (even with a different name) would be very useful indeed. Especially of course since GCC has dropped support for IRIX, so that's only going to steadily grow further out of date over time. Of course, I assume it's still capable of generating decent MIPS4 code...

Edit: Reading jan-jaaps post (added whilst I wrote this), might have guessed there's a bit more to it...

I still have no idea what you want with an updated MipsPro. There is no such a thing like a miracle compiler that makes run current open source on an O2 as fast as on any up to date hardware. MipsPro builds in many situations faster code than gcc, but we are talking about 10% or maybe 20%. We have to face the reality: the current iphone is faster than any Tezro and there will never be any compiler that will change this.

If we want to port current open source to Irix we need a lot of programming skills, because this code is build for Linux and possibly BSD. The folks that wrote this code has no idea that Irix ever exists. And there will never be a miracle compiler, that will port this code for us.

I'm around to help test...

For the past few months, I fully use an SGI for day to day work, and after work. I'm already forgetting how to use Windows and/or MacOSX and their apps. The macs and PCs are turned off, save a few VMs to RDP into for some specific webbrowsing or using Office tools/outlook required by my work. The Tezro was too loud so I ended up quieting an O2 r5k 300mhz with a fan replacement and have been happy with it. It prints fine, it has a scanner, outputs sound to a receiver for some good audio, and a video input card where I can have microscopes or stereoscopes displaying on the desktop (hobby), or even an iphone playing youtube or netflix through the video-in.

I'm a systems engineer / IT admin-type by trade. SSH, VNC, Rdesktop are my main tools. A Vmware ESX farm is always a click away with all sorts of Operating Systems available. Daily work involves diagramming/UML, documentation, spreadsheets, even 3d-modeling helps with creating presentations (3d-network designs, etc..). Ruby and Python.. as well as some Tcl/TK are also part of my toolset with IRIX. Cybersecurity tasks using python scripts. IRIX has been a dream for me to use daily even today in 2015. IRIX is about getting down to business... with a side of classic games.

Any way we can get a newer java JDK? That would open up a new world of possibilities with applications for IRIX! OpenJDK, etc..

Dia... I agree we need a newer version.

Any way to open/convert .docx and .xlsx formats would be a big bonus for Irix. For me.. AbiWord, and Prophet, for tables/spreadsheets, work well (and Framemaker and Wordperfect are rock solid too).. but nothing with docx/xlsx that the rest of the world seems to expect.

Another spreadsheet application? At least NeXTStep had Lotus Improv

Mplayer.. thanks for all the hard work that was put into this recently! For youtube videos, I still have to use youtube-dl to get 3gp format though (low quality but usually watchable), mp4 videos still give me a black screen most of the time.

Ruby.. I could use a more recent version. Also.. I am never able to install Ruby gems that have native code (compile on IRIX). If I could get Ruby with the "rvc" gem.. I could manage Vmware ESX/vsphere systems from IRIX on the command line! Plus, with Ruby I can create webapps.. manipulate LDAP or Database servers, etc.

Thunderbird is still so sluggish (it is better on a Tezro though). Slypheed is super fast.. but the current nekoware version does not support SPAM filtering. I tried compiling several newer versions.. they all compile fine but usually core dump opening any Inbox (but any other IMAP folder works fine). Newer versions of Sylpheed support SPAM filtering.. and sylfilter (for spam filtering) compiles cleanly on IRIX (need to compile GDBM first and use that and disable using sqlite.. an error with nekoware's sqlite occurs if you dont.)

Is a newer version of Blender possible? Our version is getting a bit outdated.. where it cannot open newer files very well.

transmission-bt .. or any torrent client that supports UDP trackers?

GNUCash ? Would save me from SoftWindows95 and an old version of Quicken.

Another game to see compiling on Irix could be Warzone 2100. The earlier versions need some OpenAL for 3D sound that I couldnt get to compile. Newer versions needed newer libs than what nekoware has available.
[EDIT: OpenAL works fine from nekoware ]

-Kevin

necron2600 wrote: Any way to open/convert .docx and .xlsx formats would be a big bonus for Irix ... but nothing with docx/xlsx that the rest of the world seems to expect.

Getting a little off dexter1's topic here but a few things miight help you ... I've hunted everywhere for a docx-doc converter. To me, this is a big indication of the change in computing - in the old days someone would have coded up a commandline converter. Now there's nothing.

Online converter

Another spreadsheet application?

I did find the source for Axene, which is

1) Motif
2) supposedly reads / writes Excel spreadsheets

I'd love to see the Axene programs run in Irix.

There is SIAG somewhere here - it is supposed to read xls files but the program is more than somewhat goofy.

mp4 videos still give me a black screen most of the time.

Must be the O2 All the mp4's I tried on the O350 were fine, haven't tried too many on the Octane yet. Most videos so far on the Octane seem to be okay. Agreed, neko MPlayer is a highlight. If we can get dexter1 and axatax back on that train, it'd be a winnah ! winnah !

Thunderbird is still so sluggish ...

I use Tbird only because I have so much mail already in it and too lazy to change. Not too slow on an Octane. Ishmail runs great tho and it's more Irrixy. You might try that.

GNUCash ? Would save me from SoftWindows95 and an old version of Quicken.

LedgerSMB runs in Irix. Might be overkill tho

necron2600 wrote: with Ruby I can create webapps.. manipulate LDAP or Database servers, etc.

you can do that with php

rosehillbob wrote: What this means right now people are basically running an updated version of our MipsPro compiler under MIPS Linux!
I think we should get the compiler working back under SGI!

Oh, LLVM eh? I wasn't notified I'll have to clear it...

rosehillbob wrote: What this means right now people are basically running an updated version of our MipsPro compiler under MIPS Linux!
I think we should get the compiler working back under SGI!

Uh, so would that be this? Link: https://github.com/penberg/linux-kvm/tree/master/tools/kvm

Looks like it's got a long way to go to be a replacement for MIPSPro.

Update-wise I did some more hacking on Maxwell and got it to where it compiles, but their goofy-ass build system doesn't create the executable until you do "make install," at which point it fails miserably on some kind of weird dependency check. Still hacking...

Because of the recent development posts in Maxwell, I have split the topic. Maxwell development is now in viewtopic.php?f=15&t=16730190

hamei wrote: I use Tbird only because I have so much mail already in it

ehrmmm... IMAP not available in the Middle Kingdom?