tried to configure

[jmzaleski]

I just tried to configure a new clone of ivtools-1.2 on a linuxmint 17.3

Linux matz-vbox-64-moz 3.19.0-32-generic #37~14.04.1-Ubuntu SMP Thu Oct 22 09:41:40 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

The configure complains:

writing constants to be pulled in by each Makefile from config/config.mk:
writing "XCONFIGDIR = /usr/lib/X11/X11/config"
writing "ABSTOP = /home/matz/git/ivtools-1.2"
make: *** No rule to make target `CPU'.  Stop.
writing "CPU = "
make: *** No rule to make target `CPU'.  Stop.
writing "LIBSTDCPLUSPLUS2 = 0"
writing "LIBSTDCPLUSPLUS3 = 1"

now do a "make"

but no Makefile was created.

Sorry, for the dumb question, but when ./configure --help | grep -i cpu returns nothing, I don't know what to do next.

[vectaport]

Matthew,

Thanks for your question. You have run into a problem I created while moving to a git distribution of the source. Since it is an imake based build system, I put all the Makefile’s into .gitignore. But you need an initial top-level Makefile to bootstrap (so it can do “make CPU”).

My inelegant work-around (that will need better documentation) is to add a Makefile.orig to the repository, then for now you have to manually copy it to Makefile before running configure.

So I’ve pushed this Makefile.orig. I hope this works. I will look at hacking the configure script to automate this.

Scott Johnston

On Feb 8, 2016, at 4:22 PM, Mathew Zaleski notifications@github.com wrote:

I just tried to configure a new clone of ivtools-1.2 on a linuxmint 17.3

Linux matz-vbox-64-moz 3.19.0-32-generic #37~14.04.1-Ubuntu SMP Thu Oct 22 09:41:40 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

The configure complains:

writing constants to be pulled in by each Makefile from config/config.mk: writing "XCONFIGDIR = /usr/lib/X11/X11/config" writing "ABSTOP = /home/matz/git/ivtools-1.2" make: * No rule to make target `CPU'. Stop. writing "CPU = " make: * No rule to make target`CPU'. Stop. writing "LIBSTDCPLUSPLUS2 = 0" writing "LIBSTDCPLUSPLUS3 = 1" now do a "make"

but no Makefile was created.

Sorry, for the dumb question, but when ./configure --help | grep -i cpu returns nothing, I don't know what to do next.

— Reply to this email directly or view it on GitHub https://github.com/vectaport/ivtools-1.2/issues/3.

[jmzaleski]

I did ln -s Makefile.orig Makefile ./configure make

This got a lot further, but when I went and looked for specific a.out's there weren't any. For instance, src/dclock/LINUX failed to make due a few symbols in libXext So, just hacking around, I added -L/usr/lib/x86_64-linux-gnu to the link command line and success.. More hints? Sounds like a configure option to me (?)

[vectaport]

—x-libraries=path and —x-includes=path should help. See ./configure —help for more options.

Scott

On Feb 8, 2016, at 7:25 PM, Mathew Zaleski notifications@github.com wrote:

I did ln -s Makefile.orig Makefile ./configure make

This got a lot further, but when I went and looked for specific a.out's there weren't any. For instance, src/dclock/LINUX failed to make due a few symbols in libXext So, just hacking around, I added -L/usr/lib/x86_64-linux-gnu to the link command line and success.. More hints? Sounds like a configure option to me (?)

— Reply to this email directly or view it on GitHub https://github.com/vectaport/ivtools-1.2/issues/3#issuecomment-181689109.

[barak]

Does the autotools-based system that ditches imake, in my fork of the repo, work for you?

[jmzaleski]

yes, the makefiles generated by configure work, at least as far as the link edit. (tell the truth I had forgotten all about imake. long time since fooling with IV)

I don't think --x-libraries path is my issue --- I see rules in the configure script checking if x programs will link without further -L options.

Rather, I think it's something to do with the goofy place linuxmint/ubunto put libXext.so.

I can't see anything in the configure script to add an extra -L (..but i bet it's there somewhere..)

[jmzaleski]

As usual, configure and friends have defeated me. I went on a campaign to make dclock work. In the generated Makefile in ivtools-1.2/src/dclock/LINUX/Makefile I found a macro called XLIBDIR. It's undefined (I verified this by adding echo $(XLIBDIR))

So, I defined it like so: XLIBDIR = -L/usr/lib/x86_64-linux-gnu (I know this will be overwritten, but I did it for experimentation's sake) sure enough, now the link errors complaining about symbols from libXext go away. What's more, if I do env LD_LIBRARY_PATH=$HOME/git/ivtools-1.2/src/IV/LINUX ./a.out it runs!

However, trying to figure out how to get the configure program to set XLIBDIR is beyond me.

[vectaport]

So "./configure —x-libraries=libpath” didn’t work for you? Followed by a make clean for safety before a make?

Scott

On Feb 9, 2016, at 2:06 PM, Mathew Zaleski notifications@github.com wrote:

As usual, configure and friends have defeated me. I went on a campaign to make dclock work. In the generated Makefile in ivtools-1.2/src/dclock/LINUX/Makefile I found a macro called XLIBDIR. It's undefined (I verified this by adding echo $(XLIBDIR))

So, I defined it like so: XLIBDIR = -L/usr/lib/x86_64-linux-gnu

(I know this will be overwritten, but I did it for experimentation's sake) sure enough, now the link errors complaining about symbols from libXext go away. What's more, if I do env LD_LIBRARY_PATH=$HOME/git/ivtools-1.2/src/IV/LINUX ./a.out

it runs!

However, trying to figure out how to get the configure program to set XLIBDIR is beyond me.

— Reply to this email directly or view it on GitHub https://github.com/vectaport/ivtools-1.2/issues/3#issuecomment-182099138.

[jmzaleski]

ugh. sorry. i must have made a mistake when it tried it earlier. I tried again like so:

 ./configure --x-libraries=/usr/lib/x86_64-linux-gnu
make
env LD_LIBRARY_PATH=$HOME/git/ivtools-1.2/src/IV/LINUX ./src/dclock/LINUX/a.out

works!

[jmzaleski]

Okay, so --x-libraries=/usr/lib/x86_64-linux-gnu works.. but shouldn't configure tell us it's missing essential stuff??

So, I stared at configure for a while. No surprise, I've very confused about what it's doing in its attempts to configure X. If one of you wants to give me a hint (or two) I'll try and figure out how to fix this so configure does something more helpful, like report that it doesn't know where the X libs and what to do. Well, any suggestions would be appreciated. Following are my notes as I wandered about trying to figure out what was going one.

configure, around line 3556 does:

  cat >Imakefile <<'_ACEOF'
acfindx:
    @echo 'ac_im_incroot="${INCROOT}"; ac_im_usrlibdir="${USRLIBDIR}"; ac_im_libdir="${LIBDIR}"'
_ACEOF

#following generates Makefile from Imakefile

  if (xmkmf) >/dev/null 2>/dev/null && test -f Makefile; then

#following evaluates  ac_im_incroot="${INCROOT}"; ac_im_usrlibdir="${USRLIBDIR}"; ac_im_libdir="${LIBDIR}"'
#and now we know what xmkmf had in mind for us

    eval `${MAKE-make} acfindx 2>/dev/null | grep -v make`
    # Open Windows xmkmf reportedly sets LIBDIR instead of USRLIBDIR.

#now it looks for $(USRLIBDIR)/libX11.so (etc) 

    for ac_extension in a so sl; do
      if test ! -f $ac_im_usrlibdir/libX11.$ac_extension &&
     test -f $ac_im_libdir/libX11.$ac_extension; then
    ac_im_usrlibdir=$ac_im_libdir; echo ac_im_usrlibdir $ac_im_usrlibdir; break
      fi
    done

##and never finds any of these.. isn't this a problem?? so it continues using the values it got from
## the generated makefile. and hence my link edit fails because that's not where the X libs actually are.

[vectaport]

The way the configure script reports the missing X library directory is when it prints out XCONFIGDIR at the end. If this is empty you’ve got a problem. That script could be modified to detect an empty X library variable and suggest using —x-libraries.

I have never quite figured out how to extend the search list for X library directory (the X includes directory is a pretty obvious table within the script).

Scott

On Feb 11, 2016, at 1:17 PM, Mathew Zaleski notifications@github.com wrote:

Okay, so --x-libraries=/usr/lib/x86_64-linux-gnu works.. but shouldn't configure tell us it's missing essential stuff??

So, I stared at configure for a while. No surprise, I've very confused about what it's doing in its attempts to configure X. If one of you wants to give me a hint (or two) I'll try and figure out how to fix this so configure does something more helpful, like report that it doesn't know where the X libs and what to do. Well, any suggestions would be appreciated. Following are my notes as I wandered about trying to figure out what was going one.

configure, around line 3556 does:

cat >Imakefile <<'_ACEOF' acfindx: @echo 'ac_im_incroot="${INCROOT}"; ac_im_usrlibdir="${USRLIBDIR}"; ac_im_libdir="${LIBDIR}"' _ACEOF

following generates Makefile from Imakefile

if (xmkmf) >/dev/null 2>/dev/null && test -f Makefile; then

following evaluates ac_im_incroot="${INCROOT}"; ac_im_usrlibdir="${USRLIBDIR}"; ac_im_libdir="${LIBDIR}"'

and now we know what xmkmf had in mind for us

eval `${MAKE-make} acfindx 2>/dev/null | grep -v make`
# Open Windows xmkmf reportedly sets LIBDIR instead of USRLIBDIR.

now it looks for $(USRLIBDIR)/libX11.so (etc)

for ac_extension in a so sl; do
  if test ! -f $ac_im_usrlibdir/libX11.$ac_extension &&
 test -f $ac_im_libdir/libX11.$ac_extension; then
ac_im_usrlibdir=$ac_im_libdir; echo ac_im_usrlibdir $ac_im_usrlibdir; break
  fi
done

and never finds any of these.. isn't this a problem?? so it continues using the values it got from

the generated makefile. and hence my link edit fails because that's not where the X libs actually are.

— Reply to this email directly or view it on GitHub https://github.com/vectaport/ivtools-1.2/issues/3#issuecomment-183063010.

[vectaport]

To summarize there are 3 common problems with using the configure script and imake to build ivtools from source:

1) configure script cannot find X libraries -- solution is to provide --x-libraries argument 2) make CPU doesn't come up with the right CPU value to name the directories that object code and binaries are in -- solution is to set CPU environment variable yourself, either DARWIN for MacOS or LINUX for everything else nowadays. 3) there is no top-level Makefile if built from clone GitHub -- solution is to manually copy Makefile.orig to Makefile before doing the first make after running configure.