cygwin64 build error for ppl-1.1

[Sudarsan Vasista Srinivasan Ranganathan]

Hi,

I'm trying to install Sage on cygwin64 with the operating system being Windows 8.1. I know of the persisting issues along with remedies posted at http://trac.sagemath.org/wiki/Cygwin64Port#TestingSage5.13onWindows764bitsrunningwithinVirtualBox4.3withCygwin64 but even with the the weak.patch fix #16152 for ppl-1.1 my build fails for reasons that I cannot comprehend.

The log file of the build can be found here: https://dl.dropboxusercontent.com/u/93147662/ppl-1.1.log

I'd appreciate if help can be provided on this topic.

P. S. I installed mpir-2.7.0.alpha4 which was the only tweak that I've used so far before this build patch for ppl1.1.

[Dima Pasechnik]

On 2014-08-15, Sudarsan Vasista Srinivasan Ranganathan <sudarsa...@gmail.com> wrote:
> ------=_Part_958_766231161.1408137021885
> Content-Type: text/plain; charset=UTF-8

>
> Hi,
>
> I'm trying to install Sage on cygwin64 with the operating system being
> Windows 8.1. I know of the persisting issues along with remedies posted at
> http://trac.sagemath.org/wiki/Cygwin64Port#TestingSage5.13onWindows764bitsrunningwithinVirtualBox4.3withCygwin64
> but even with the the weak.patch fix #16152

><http://trac.sagemath.org/ticket/16152> for ppl-1.1 my build fails for

> reasons that I cannot comprehend.
>
> The log file of the build can be found here:
> https://dl.dropboxusercontent.com/u/93147662/ppl-1.1.log
>
> I'd appreciate if help can be provided on this topic.
>
> P. S. I installed mpir-2.7.0.alpha4 which was the only tweak that I've used
> so far before this build patch for ppl1.1.

As the errors you get are related to gmp, which is meant to be provided by mpir,
perhaps that is the reason.

What is the compiler you are using? The one from Cygwin, or the one bundled with Sage?


>

[Sudarsan Vasista Srinivasan Ranganathan]

I'm not setting any flags prior to running the makefile for Sage; so safe to presume its the one from Cygwin? My Cygwin has GCC 4.8.3-2.

[Dima Pasechnik]

On 2014-08-16, Sudarsan Vasista Srinivasan Ranganathan <sudarsa...@gmail.com> wrote:
> ------=_Part_1311_1972416071.1408178920371
> Content-Type: text/plain; charset=UTF-8

>
> I'm not setting any flags prior to running the makefile for Sage; so safe
> to presume its the one from Cygwin? My Cygwin has GCC 4.8.3-2.

before deciding whether to build gcc, or to use the one installed, Sage does some checks.
Not sure exactly what. You may try running Sage shell
(./sage -sh)
and at its prompt type
gcc -v

Tha latter will show the compiler version used by Sage.

By the way, how did you istall that MPIR alpha4?

>
> On Saturday, August 16, 2014 1:12:19 AM UTC-7, Dima Pasechnik wrote:
>>
>> On 2014-08-15, Sudarsan Vasista Srinivasan Ranganathan <

[Sudarsan Vasista Srinivasan Ranganathan]

On Saturday, August 16, 2014 5:52:40 AM UTC-7, Dima Pasechnik wrote:

On 2014-08-16, Sudarsan Vasista Srinivasan Ranganathan <sudarsa...@gmail.com> wrote:
> ------=_Part_1311_1972416071.1408178920371
> Content-Type: text/plain; charset=UTF-8
>
> I'm not setting any flags prior to running the makefile for Sage; so safe
> to presume its the one from Cygwin? My Cygwin has GCC 4.8.3-2.

before deciding whether to build gcc, or to use the one installed, Sage does some checks.
Not sure exactly what. You may try running Sage shell
(./sage -sh)
and at its prompt type
gcc -v

Tha latter will show the compiler version used by Sage.

I ran the command and it says GCC 4.8.3 - the same as what I've installed with my Cygwin64. I guess the sage source that I have has a GCC version 4.7.3 packed as a tarball in the upstream folder. 

By the way, how did you istall that MPIR alpha4?

I synced the folder /build/pkgs/mpir to the files over here: http://git.sagemath.org/sage.git/commit/?id=a564118a655d3af6730c103ba18ef6f7fb0cb397  - this is ticket #15015; and then I downloaded the tarball for the version from the ticket's homepage and let sage build itself. The reason why I chose to go with this is that the patches for mpir 2.6.0.p4 at various tickets didn't work for me because I was making a mistake while syncing the patch files - I realized this only later. Do you think that removing mpir's latest 2.7.0alpha and reverting back to 2.6.0p4 will make ppl work along with its patch? Or is there something that i can do now?

[Sudarsan Vasista Srinivasan Ranganathan]

Update: It was the mpir2.7.0 alpha that caused problems while installing ppl - the patched ppl with mpir2.6.0.p4 works; I reverted back to this. The build continues.....

[Sudarsan Vasista Srinivasan Ranganathan]

Updated (new) question regarding the build process:

Now I seem to be stuck with ATLAS that has been taking more than 7 hours to build as of now and it's still going on. Is this even supposed to happen? Am I doing something wrong?

On Friday, August 15, 2014 2:10:22 PM UTC-7, Sudarsan Vasista Srinivasan Ranganathan wrote:

[Sudarsan Vasista Srinivasan Ranganathan]

Update: ATLAS did build and here are the stats:

real    531m45.862s
user    236m20.584s
sys     96m25.287s
Successfully installed atlas-3.10.1.20140210

I don't know why it took so long - any other user faced a similar situation before?

On Friday, August 15, 2014 2:10:22 PM UTC-7, Sudarsan Vasista Srinivasan Ranganathan wrote:

[Dima Pasechnik]

On 2014-08-17, Sudarsan Vasista Srinivasan Ranganathan <sudarsa...@gmail.com> wrote:
> ------=_Part_788_2049542337.1408299073663
> Content-Type: text/plain; charset=UTF-8

>
> Update: ATLAS did build and here are the stats:
>
> real 531m45.862s
> user 236m20.584s
> sys 96m25.287s
> Successfully installed atlas-3.10.1.20140210
>
> I don't know why it took so long - any other user faced a similar situation
> before?

building Altas includes tuning, and it can be slow - and Cygwin is slowing it
down quite a bit. Yes, 7 hours is not exceptionally long.

Are you doing a parallel build?


One way to avoid this is to use system-wide Atlas.
(see SAGE_ATLAS_LIB in
"Environment variables" section of the installation manual)

>
> On Friday, August 15, 2014 2:10:22 PM UTC-7, Sudarsan Vasista Srinivasan
> Ranganathan wrote:
>>
>> Hi,
>>
>> I'm trying to install Sage on cygwin64 with the operating system being
>> Windows 8.1. I know of the persisting issues along with remedies posted at
>> http://trac.sagemath.org/wiki/Cygwin64Port#TestingSage5.13onWindows764bitsrunningwithinVirtualBox4.3withCygwin64
>> but even with the the weak.patch fix #16152

>> <http://trac.sagemath.org/ticket/16152> for ppl-1.1 my build fails for

[Sudarsan Vasista Srinivasan Ranganathan]

On Sunday, August 17, 2014 11:27:37 AM UTC-7, Dima Pasechnik wrote:

On 2014-08-17, Sudarsan Vasista Srinivasan Ranganathan <sudarsa...@gmail.com> wrote:
> ------=_Part_788_2049542337.1408299073663
> Content-Type: text/plain; charset=UTF-8
>
> Update: ATLAS did build and here are the stats:
>
> real    531m45.862s
> user    236m20.584s
> sys     96m25.287s
> Successfully installed atlas-3.10.1.20140210
>
> I don't know why it took so long - any other user faced a similar situation
> before?

building Altas includes tuning, and it can be slow - and Cygwin is slowing it
down quite a bit. Yes, 7 hours is not exceptionally long.

Are you doing a parallel build?

No I'm not; I'm only doing a serial build.
 

One way to avoid this is to use system-wide Atlas.
(see SAGE_ATLAS_LIB in
"Environment variables" section of the installation manual)

Thanks for this info - but now that it has built itself I don't need this right?

Update:
On an entirely different note I now seem to have run into the next problem; it's ECL now - you can find my log here: https://dl.dropboxusercontent.com/u/93147662/ecl-12.12.1.20140409.log

I understand why it's failing - seems to be a missing file error but I cannot understand how or why the build process is running into this issue. Any help would be appreciated on this. Could this be related to the problems that ECL might encounter when building in a subdirectory of a path that has a folder starting with capital letter somewhere? For instance, my cygwin64 username is "Sudarsan" but nothing else in the path has its first letter as a capital letter.
 

[Jean-Pierre Flori]

If you want to attempt a build on Cygwin64 pleas follow http://trac.sagemath.org/wiki/Cygwin64Port and use the u/jpflori/cygwin64 branch on Sage's git.
(Unless that's already what you are doing.)

[Sudarsan Vasista Srinivasan Ranganathan]

I did not use that branch but I have been following the instructions particularly in the last two posts. Does the error in my last log related to ecl mean that I should have probably used the branch that you're pointing to?

[Sudarsan Vasista Srinivasan Ranganathan]

The build seems to be failing because the installer cannot find ../build/bin/ecl.exe and even I cannot see that file in the directory. Can't understand really as to why the file is missing in the first place.

On Friday, August 15, 2014 2:10:22 PM UTC-7, Sudarsan Vasista Srinivasan Ranganathan wrote:

[Sudarsan Vasista Srinivasan Ranganathan]

*Bump*. Any thoughts on my above weird "missing file" error?

On Friday, August 15, 2014 2:10:22 PM UTC-7, Sudarsan Vasista Srinivasan Ranganathan wrote:

[Jean-Pierre Flori]

As far as ATLAS is concerned, it's hellishly long and that's to be expected.
You can try using a pretuned config by setting SAGE_ATLAS_ARCH=base/fast or using SAGE_ATLAS_LIB to use system-wide libraries as Dima suggested.

I'll have a look at ECL issue tonight.

[Jean-Pierre Flori]

The ECL issue looks strange...
The failed install should have left the build directory somewhere in Sage's tree (look under local/var/tmp/.../ecl-.../ or smtg like that).
Can you cd there and check that ecl.exe is actually not present?
(I'll try to give you more details later on the directory names and so on if you don't manage them by yourself.)

In fact I guess it won't be there.
It seems the build process only built the minimal lisp interpreter/compiler and then jump to the end of the install process...
The log on a working system is much longer.
In particular it should be full of lines like:

(compile-file "src:lsp;export.lsp" :output-file #P"BUILD:LSP;EXPORT.O.NEWEST" :SYSTEM-P T :C-FILE T :DATA-FILE T :H-FILE T)
triggering compilation of lisp files just after
;;; End of Pass 1.sed -e 's,@libdir\\@,/home/Sudarsan/sage/sage-6.3/local/lib/,' \
    -e 's,@includedir\\@,/home/Sudarsan/sage/sage-6.3/local/include/,' \
    -e 's,~A,/home/Sudarsan/sage/sage-6.3/local/lib/,' bin/ecl-config.pre > bin/ecl-config
The sed stuff here is fishy...
There's even no newline.

It should read:
;;; End of Pass 1.
;;; Emitting code for DOLIST.
;;; Emitting code for DOTIMES.
;;; Emitting code for DO/DO*-EXPAND.
;;; Emitting code for EVAL-FEATURE.
;;; Emitting code for DO-READ-FEATURE.
;;; Emitting code for SHARP-+-READER.
;;; Emitting code for SHARP---READER.
;;; Note:
;;;   Invoking external command:
;;;   gcc -I. -I/home/jpflori/sage.git/local/var/tmp/sage/build/ecl-12.12.1.20140409/src/build/ -DECL_API -I/home/jpflori/sage.git/local/var/tmp/sage/build/ecl-12.12.1.20140409/src/build/c -I/home/jpflori/sage.git/local/include -I/home/jpflori/sage.git/local/include -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -g -O2 -fPIC -Dlinux -I/home/jpflori/sage.git/local/var/tmp/sage/build/ecl-12.12.1.20140409/src/src/c -w -c lsp/export.c -o lsp/export.o
;;; Finished compiling SRC:LSP;EXPORT.LSP.
;;;
(compile-file "src:lsp;defmacro.lsp" :output-file #P"BUILD:LSP;DEFMACRO.O.NEWEST" :SYSTEM-P T :C-FILE T :DATA-FILE T :H-FILE T)
;;;
;;; Compiling SRC:LSP;DEFMACRO.LSP.
;;; OPTIMIZE levels: Safety=2, Space=1, Speed=1, Debug=1
blabablabalblabalbalbalbalbalbalbalb
-> installation success

Can you try the cygwin64 branch?
I'll also try to update it to the latest stable version of Sage (plus needed patches; hint: review the corresponding tickets) before end of August.

[Sudarsan Vasista Srinivasan Ranganathan]

Hi,

Thanks for the reply. I'll surely do try building sage from your cygwin64 branch. And I'm not sure if I will get the time to do the reviews by the end of August (wrapping up some work and thus the delayed post). I haven't gone back and looked at my own build issue either. I'm new to both the trac and the developer forum - is there a release by the end of August?