IRC channel logs

***sneek_ is now known as sneek

<civodul>Hello Guilers!

<stis>hi civodul:

<bhattigurjot>does this "GUILE_CFLAGS = @GUILE_CFLAGS@" work in guile-2.0?

<bhattigurjot>in the make file I mean?

<dsmith-work>bhattigurjot: Not in a Makefile, but in the Makefile.in. Usually ./configure does something with sed to change @...@ markers to the real conigured values.

<bhattigurjot>yeah in that.

<dsmith-work>IF you see @...@ in a Makefile, it's an indication that ./configured didn't do the right thign.

<bhattigurjot>dsmith-work: can you take a look at this file

<bhattigurjot> http://git.savannah.gnu.org/cgit/dr-geo.git/tree/Makefile.am?h=upgrade

<bhattigurjot>is it not correct way to include GUILE_CFLAGS ?

<bhattigurjot>this is defined in makefile.in as "GUILE_CFLAGS = @GUILE_CFLAGS@"

<bhattigurjot>is it incorrect to include it in AM_CPPFLAGS ?

<dsmith-work>Here is the way I did it: https://www.gitorious.org/guile-sqlite/guile-sqlite/source/54acbdb2efef8a9b62892b92c97dfef2024810c0:

<bhattigurjot>dsmith-work: you also used a variable to include GUILE_CFLAGS

<bhattigurjot>value

<bhattigurjot>and I did too. I don't know why is it not working

<bhattigurjot>:/

<dsmith-work>Looks like "PKG_CHECK_MODULES([GUILE], [guile-2.0])" generates that.

<dsmith-work>Anyone know how to make that work for 2.2 (2.1) also?

<dsmith-work>bhattigurjot: btw, I need to set PKG_CONFIG_PATH. As in "PKG_CONFIG_PATH=/usr/local/lib/pkgconfig ./configure"

<bhattigurjot>dsmith-work: I can see that the Makefile has the value

<bhattigurjot>GUILE_CFLAGS = -pthread -I/usr/include/guile/2.0

<bhattigurjot>and do you know what should be used inplace of "gh_scm2doubles (v, gv);"?

<bhattigurjot>I mean this syntax was used in gh interface

<dsmith-work>bhattigurjot: I'm using pkg-config directly. Not using the /usr/local/share/aclocal/guile.m4

<bhattigurjot>what is scm interface replacement?

<dsmith-work>bhattigurjot: The NEWS file has all that

<dsmith-work>scm_to_double() probably

<dsmith-work>IT's a much cleaner, saner interface

<bhattigurjot>actually scm_to_double takes a single parameter

<bhattigurjot>dsmith-work: they said to use Uniform Numeric Vector

<bhattigurjot>that's where I am confused

<dsmith-work>bhattigurjot: Sorry, no idea.

<dsmith-work>civodul: Do you know?

<civodul>dsmith-work, bhattigurjot: scm_to_double, yes

<civodul>ah no

<civodul>bhattigurjot: you want to get a vector of doubles?

<bhattigurjot>civodul: here is the code

<bhattigurjot> http://pastebin.com/R7kq2pzM

<civodul>scm_take_f64vector would give a homogeneous vector

<civodul>bhattigurjot: i don't know what gh_scm2doubles did, hence my question :-)

<dsmith-work>bhattigurjot: Sounds like you are trying to port some really really old guile code.

<bhattigurjot>yeah

<bhattigurjot>very old

<bhattigurjot>civodul: here is what I think, https://www.gnu.org/software/guile/docs/docs-1.8/guile-ref/Scheme-to-C.html

<dsmith-work>It's some work, but it might be instructive to get the source to a guile where the gh_* was in the depracated state. There are usually files that map the old api to the new one. Useful for clues on how to do it.

<civodul>bhattigurjot: the closest thing would be to use SRFI-4 "homogeneous vectors" instead of plain vectors, and then scm_f32vector_elements to access the vector's elements as a C array

<bhattigurjot>take a look here

<bhattigurjot> http://pastebin.com/ev0HPv5N

<bhattigurjot>I did that using array length

<bhattigurjot>it worked but I must use guile 2.0.11 for that

<civodul>bhattigurjot: what you did may work, but it's inefficient

<civodul>SRFI-4 homogeneous vectors were designed for these use cases

<civodul>where you have large arrays of numbers to be handled both from C and Scheme

<bhattigurjot>civodul: can you give me an example

<civodul>i don't have examples handy, and i'm a bit busy ;-)

<civodul>hopefully http://www.gnu.org/software/guile/manual/html_node/SRFI_002d4.html gives a good introduction

<bhattigurjot>ah... thanks btw :-)

<dsmith-work>civodul: Looks that gh_scm2doubles never went through the deperecation cycle.

<dsmith-work>civodul: It's in 1.8 (not deprecated) and disappears in 2.0

<civodul>oh, i think all of gh did, no?

<civodul>not deprecated?

<civodul>all of gh was deprecated AFAIK

<dsmith-work>Hmm. Maybe you are right.

<dsmith-work>NEWS never mentions a replacement.

<civodul>oh

<civodul>good night/day!

IRC channel logs

2014-06-09.log