Re: Issues & questions building Octave-3.4.2 MinGW

[Tatsuro MATSUOKA]

Hello

--- On Sun, 2011/7/10, Philip Nienhuis  wrote:

> Hi Tatsuro:
> 
> Tatsuro MATSUOKA wrote:
> > Hello
> > 
> > --- On Sun, 2011/7/10, Philip Nienhuis  wrote:
> > .
> >> I also found a number of "gpnewlib" entries (=>  gplibs) and stray
> >> include dirs starting with "/projetos/gcc/ming/..."
> >> 
> >> <snip>
> > 
> > Oops!  I have missed them when I renamed the directory for gnuplot.
> > I will check it and modify them and upload corrected library archives.
> 
> Are you sure a new upload is really needed?
> 
> First, Octave does get built OK in spite of these "wrong" path settings.
> The only issue is that tons and tons of libtool messages are generated during 
> "make" about "lib<blah-blah> seems to be moved" - AFAICS unnecessary and even 
> annoying as the configure script has found those libs anyway and Octave 
> passes almost all tests.
> 
> Second, your path settings are only valid if people install your build 
> environment at exactly the same location as where you have put them.
> I didn't do that - C:\ is my system partition and I never do development work 
> there - the "conftest error" (see below) is a fine example why. So some 
> people need to adapt (all these) path settings all the same.

OK, I will correct them for myself.  If  really update the dependency 
libraries, I will update them.
Thanks for your suggestion. 
> As to the tests, the three unexpected FAILs are in:
> - imread.m
> "imread: invalid image file: Magick++ exception: Magick: Unable to access 
> configuration file (magic.mgk) reported by magick/blob.c:1830 
> (GetConfigureBlob)"

Hmmm. I have never seen this error.

> - data.cc
> where NaN / NaNi are returned rather than the expected Inf / Infi.
> 

Perhaps, this problem has been known from octave 3.0.
The treatments of NaN on MinGW-gcc seems to be different from those on other 
plat forms.

 

> My pleasure. While trying to adapt a few things I'm slowly getting to 
> understand bit by bit what you have accomplished and how the whole stuff 
> sticks together. I knew, and can confirm, that it must have been a lot of 
> work.
> Thank *you* again for sharing the results.
> 
> Do you also have a description of how you have set up the Octave-3.4.2 
> directory (the "binary") itself?

I did not make documentation for my work adding dll files etcs. in addition to 
stuffs installed by make install-strip.
I will consider make a note for this work.
Please wait a while.


> 
> >>> On. WIndows XP. It alwas occur in checking hard link.
> >>> However on windows 7, it has not been occurred.
> >>> Perhaps checking routine should be modified for windows XP but I cannot 
> >>> find the way to correct it.
> >> 
> >> Indeed, it happens when configure tests if rename manages hard links.
> >> I'll file a bug report for this.
> 
> Done.
I have confirmed that you have registered the issue on the bug tracker.
I made cc. of the mail to me if someone commit this issue.
> 
> Me too. This situation is fairly discouraging - not only for developers:
> my 8-yr old daughter has a hard time understanding why she has to wait half 
> an hour before she can use the computer after I'm done :-)

Logically it is possible to correct this check for MinGW platform as you wrote 
in the bug tracker.  I hope the developers will accept your proposal
"Or could a configure option "--enable-MinGW" or so be added to avoid this 
test? " 

For the convenience those who will read this mail, I will write the address of 
this issue on the bug tracker

https://savannah.gnu.org/bugs/?33736 

Regards

Tatsuro