Re: Successful compilation with MinGW

[Timothée Lecomte]

Ok that is what I did suspect : you have been successful too ;-)

I think you are asking the good questions, and your choices seems 
reasonable. In particular waiting for the next version of octave-forge 
specifically designed for 2.9.x is appropriate.

I also noted that running in Msys rxvt gives strange results, with in 
particular no "octave>" at the start of the line. However I also noted 
that Windows gives an error if you close with the "X" the cmd.exe 
terminal... Too bad ! I agree that an IDE would be interesting, but I 
believe that it will be hard to achieve ... This IDE will first need 
something like a terminal, which sounds like duplicating already 
existing solutions (like rxvt), but what to include next ? A script 
editor ? Probably better to add octave syntax to an already exisiting 
editor with syntax hghlighting such as Notepad++. Buttons for direct 
access to functions like gnuplot under windows ? I really think that 
these buttons are not useful at all... The last thing that come to my 
mind is something like a functions list or a graphical help system, but 
again it's probably not worth it. To conclude, it would be ideal to fix 
rxvt or to find an simple and working equivalent.

By the way, I am not using windows on a regular basis (except for skype 
video, but that's off-topic...) but I was just sensible to the fact that 
octave would greatly benefit from a good package for windows without 
cygwin, which is great but probably doesn't fit windows users' habits. 
In my school labs, most computers are running windows and ... Matlab... 
and my friends' computers too... On the other hand, I have "converted" a 
friend to gnuplot, thanks to its windows binaries.

Regards,

Timothée



PS. : To finish with my remarks about octave, I have been surprised to 
see that octave uses Makefile.in's but not their higher-level 
conterparts, Makefile.am's. Is there a reason for it ?




David Bateman a écrit :

> Yes John dropped glob/fnmatch as they are in libc on almost every 
> system and felt no need to keep them just for MinGW. I equally took 
> them from the latest glibc and compiled them separately for octave. 
> I'll distribute it as a separate library beside octave/mingw when I do 
> a release.
>
> I have a version with all of the dependencies except ATLAS (can't 
> build ATLAS under mingw, need cygwin/mingw and then can't do it under 
> vmware since the cache size can't be detected) and all of octave-forge 
> except the symbolic toolbox as CLN uses fancy templates that g++ under 
> mingw doesn't like. For the ATLAS issue, I've in fact linked octave 
> against an external version of lapack and blas, with the idea that I 
> can post-hoc replace them with ATLAS versions. ATLAS is a bastard mix 
> of blas and a few lapack  functions and so I'll need to split the 
> ATLAS build up and recreate compatible blas/lapack libraries to my 
> current external versions, but this is down fairly often under Linux 
> and is therefore a known solution. As for CLN, I think we'll be doing 
> without at least at the outset.
>
> I have every intention of doing an NSIS installer of octave/mingw in 
> the near future, but needed several points resolved
>
> 1) I wanted to release a version against a full release of octave and 
> octave-forge. I didn't want to do it against a CVS version as or a 
> version that was heavily patched as this meant that it is unclear what 
> version of the code a mingw user was using when reporting bugs.
> 2) Need an octave-forge release with the autoload to replace the 
> symbolic links (mingw doesn't have sym links and copies the file 
> instead) since as you notice octave is very large mainly due to 
> multiple copies of the same file. This just when in to octave-forge 
> and so the next release of octave-forge in the very near future 
> together with a 2.9.5 release of octave and these two conditions will 
> be met
> 3) I need to learn how to use NSIS as I've never built a windows 
> installer package..
> 4) Need to strip out some of the development libraries and headers to 
> make the package a reasonable size.
> 5) I need more hours in my day than I currently have...
>
> There are lots of little issues that you might not have noticed yet. 
> The main one being DON'T use the msys rxvt to run octave in. This 
> terminal is badly broken and readline completely messes up in an msys 
> rxvt. Your options are then a cygwin rxvt, which kind of misses the 
> point of doing a mingw build of octave in the first place, or use the 
> dos cmd.exe terminal, which is what I'm currently doing, though you'll 
> loose the capability to cut and paste. My thoughts were that this 
> mingw release of octave is a temporary measure and what is really 
> wanted under windows is a full IDE, and it that case it'll supply its 
> own terminal capabilities to integrate with octave. Sorry but it won't 
> be me to write such an IDE though, I don't even use windows, but 
> recognize that octave needs a windows native version of to get proper 
> acceptance in the windows world.
>
> Cheers
> David