reduce-algebra-developers — Discussion of development, administration and support for Reduce

[Reduce-algebra-developers] CSL on the Mac

[Arthur N. <ac...@ca...>]

It looks as if the checkin earlier today that helps on Ubuntu causes 
trouble on the Mac where the std::filesystem stuff is not supported
on the OS currently set as a minimum deployment target. At least I think
that is the trouble and I will look more tomorrow. Arthur

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Arthur N. <ac...@ca...>]

It seems that one says
   snap connections firefox
to see what permissions the CURRENT snap version has, but that does not
provide any guarantee about future releases I guess.
Based on that on ubuntu 22.04 (and who can tell what possibly different
security options will apply elsewhere or in the future) it looks to me
as if I will get away with merely arranging that the CSL code that 
launches a web browser arranges to follow a symbolic link from where it
seed reduce.doc to where the "real" version is stored.
I think that just a call of std::filesystem::canonical may achive the
guts of what I need, and if so that is a big plus for C++17.

I may try looking into the details tomorrow.

Arthur

On Tue, 3 May 2022, Rainer Schöpf wrote:

> I think the problem is that redcsl uses
>
>   file:///usr/lib/reduce/cslbuild/csl/reduce.doc/index.html
>
> to access the documentation, but this is only a link to
>
>  /usr/share/doc/reduce
>
> Access to subfolders of /usr/lib is denied, but files below /usr/share/doc can be read by Firefox.
>
> Whether this behaviour protects against malicious code or not can be argued either way. I fear that we will not be able to change it, at least not immediately.
>
> I had a quick look into the CSL source code. It seems that the help files are located relative to the program directory. It may be better to be able to configure a  directory for them, either in the configure script or as a runtime option.
>
> What do you think?
>
>   Rainer
>
>
> Am 2. Mai 2022 17:02:23 MESZ schrieb Arthur Norman <ac...@ca...>:
>> I view this as HORRIBLE and it will surely impact any other package that
>> tries to make its documentation available somewhere sensible and central such as in /usr/share and that decided to use html as a good format.
>> My reading up about snap thus far suggests that a snap package contains its own list of any places it will allow itself access to. I happen to suspect that loading .html from within my home directory is liable to be a lot more risky than anything from /usr/share, but the snap enthusiasts seem to have taken a strong view on "security".
>>
>> This hits firefox in Ubuntu today but I can see no certainty that other browsers will not end up equally "protected from harm" so merely switching the one to try feels to me as if it would risk being a short term response.
>>
>> Pretty much the only two suggestions I can come up with today (and I have
>> not checked or tested either) would be
>> (1) Ensure all the help stuff was available at http://sourceforge....
>> so that one got remote access (and were assumed to have a good network connection all the time!)
>> (2) Arrange that the first time somebody hits the "help" button that Reduce copies everything relevant into $HOME/.reduce/help-files and points firefox at stuff there.
>>
>> I fear that (2) looks safer. What do people thing. And can we have a write in campaign to both mozilla and ubuntu to the effect that this change which we had not been aware of until it hit us feels STUPID and counterproductive!
>>
>> Arthur
>>
>>
>> On Mon, 2 May 2022, Rainer Schöpf wrote:
>>
>>> I suspect that apparmor is blocking the access. There is a special apparmor
>>> profile for snapd.firefox.firefox. Anything in the system logs?
>>>
>>>  Rainer
>>>
>

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Rainer S. <rai...@gm...>]

I think the problem is that redcsl uses

   file:///usr/lib/reduce/cslbuild/csl/reduce.doc/index.html

to access the documentation, but this is only a link to

  /usr/share/doc/reduce

Access to subfolders of /usr/lib is denied, but files below /usr/share/doc can be read by Firefox.

Whether this behaviour protects against malicious code or not can be argued either way. I fear that we will not be able to change it, at least not immediately.

I had a quick look into the CSL source code. It seems that the help files are located relative to the program directory. It may be better to be able to configure a  directory for them, either in the configure script or as a runtime option.

What do you think?

   Rainer


Am 2. Mai 2022 17:02:23 MESZ schrieb Arthur Norman <ac...@ca...>:
>I view this as HORRIBLE and it will surely impact any other package that
>tries to make its documentation available somewhere sensible and central such as in /usr/share and that decided to use html as a good format.
>My reading up about snap thus far suggests that a snap package contains its own list of any places it will allow itself access to. I happen to suspect that loading .html from within my home directory is liable to be a lot more risky than anything from /usr/share, but the snap enthusiasts seem to have taken a strong view on "security".
>
>This hits firefox in Ubuntu today but I can see no certainty that other browsers will not end up equally "protected from harm" so merely switching the one to try feels to me as if it would risk being a short term response.
>
>Pretty much the only two suggestions I can come up with today (and I have
>not checked or tested either) would be
>(1) Ensure all the help stuff was available at http://sourceforge....
>so that one got remote access (and were assumed to have a good network connection all the time!)
>(2) Arrange that the first time somebody hits the "help" button that Reduce copies everything relevant into $HOME/.reduce/help-files and points firefox at stuff there.
>
>I fear that (2) looks safer. What do people thing. And can we have a write in campaign to both mozilla and ubuntu to the effect that this change which we had not been aware of until it hit us feels STUPID and counterproductive!
>
>Arthur
>
>
>On Mon, 2 May 2022, Rainer Schöpf wrote:
>
>> I suspect that apparmor is blocking the access. There is a special apparmor
>> profile for snapd.firefox.firefox. Anything in the system logs?
>> 
>>  Rainer
>> 

Re: [Reduce-algebra-developers] reduce-mode.el: bugs in syntax highlighting

[Andrey G. G. <A.G...@in...>]

Dear Francis,

for all n1,n2 match x1^n1*x2^n2=f(n1,n2);

is also highlighted incorrectly. I suppose match should have the same 
highlighting properties as let and clear.

a where {f(~x)=>0,y=>0} when x>0;

where is just black; it would be better to highlight it.
when is blue, as if it is a function; it would be better to highlight it.
I think let, match, clear, where, when should be in magenta.

Andrey

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Alan B. <ala...@gm...>]

Hi Arthur,

I agree with your comments and made much the same points to Francis .  
Fortunately snap is restricted to Linux systems for the foreseeable 
future. I can't see any glaring security problems by allowing html files 
in (say) /usr/local, /opt, /usr/share etc to be loaded (quite the 
contrary), but I'm a snap novice.

I will investigate  Rainer's suggestions re apparmor, but I am not too 
hopeful.

Experimenting I found that it is not possible to install firefox with 
'classic confinement' (which ought to do the trick, I suspect); the 
classic option  in

snap install --classic  firefox

is simply ignored as the firefox snap uses 'strict confinement'.

Alan

On 02/05/2022 16:02, Arthur Norman wrote:
> I view this as HORRIBLE and it will surely impact any other package that
> tries to make its documentation available somewhere sensible and 
> central such as in /usr/share and that decided to use html as a good 
> format.
> My reading up about snap thus far suggests that a snap package 
> contains its own list of any places it will allow itself access to. I 
> happen to suspect that loading .html from within my home directory is 
> liable to be a lot more risky than anything from /usr/share, but the 
> snap enthusiasts seem to have taken a strong view on "security".
>
> This hits firefox in Ubuntu today but I can see no certainty that 
> other browsers will not end up equally "protected from harm" so merely 
> switching the one to try feels to me as if it would risk being a short 
> term response.
>
> Pretty much the only two suggestions I can come up with today (and I have
> not checked or tested either) would be
> (1) Ensure all the help stuff was available at http://sourceforge....
> so that one got remote access (and were assumed to have a good network 
> connection all the time!)
> (2) Arrange that the first time somebody hits the "help" button that 
> Reduce copies everything relevant into $HOME/.reduce/help-files and 
> points firefox at stuff there.
>
> I fear that (2) looks safer. What do people thing. And can we have a 
> write in campaign to both mozilla and ubuntu to the effect that this 
> change which we had not been aware of until it hit us feels STUPID and 
> counterproductive!
>
> Arthur
>
>
> On Mon, 2 May 2022, Rainer Schöpf wrote:
>
>> I suspect that apparmor is blocking the access. There is a special 
>> apparmor
>> profile for snapd.firefox.firefox. Anything in the system logs?
>>
>>  Rainer
>>

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Arthur N. <ac...@ca...>]

I view this as HORRIBLE and it will surely impact any other package that
tries to make its documentation available somewhere sensible and central 
such as in /usr/share and that decided to use html as a good format.
My reading up about snap thus far suggests that a snap package contains 
its own list of any places it will allow itself access to. I happen to 
suspect that loading .html from within my home directory is liable to be a 
lot more risky than anything from /usr/share, but the snap enthusiasts 
seem to have taken a strong view on "security".

This hits firefox in Ubuntu today but I can see no certainty that other 
browsers will not end up equally "protected from harm" so merely switching 
the one to try feels to me as if it would risk being a short term 
response.

Pretty much the only two suggestions I can come up with today (and I have
not checked or tested either) would be
(1) Ensure all the help stuff was available at http://sourceforge....
so that one got remote access (and were assumed to have a good network 
connection all the time!)
(2) Arrange that the first time somebody hits the "help" button that 
Reduce copies everything relevant into $HOME/.reduce/help-files and points 
firefox at stuff there.

I fear that (2) looks safer. What do people thing. And can we have a write 
in campaign to both mozilla and ubuntu to the effect that this change 
which we had not been aware of until it hit us feels STUPID and 
counterproductive!

Arthur


On Mon, 2 May 2022, Rainer Schöpf wrote:

> I suspect that apparmor is blocking the access. There is a special apparmor
> profile for snapd.firefox.firefox. Anything in the system logs?
>
>  Rainer
>

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Rainer S. <rai...@gm...>]

Maybe it would be better to start the standard browser with

 /usr/bin/xdg-open <URL or pathname>

 Rainer


On Sun, 1 May 2022 at 17:26 +0100, Alan Barnes wrote:

 > Simply making Chrome the default browser doesn't affect which browser is used
 > by CSL GUI help. This is Firefox by default.
 >
 > To change this, one needs to "Select Browser" from the Help menu, check
 > "User" in the dialog that appears and set the command needed to launch the
 > desired browser (google-chrome,  say).
 >
 > Setting the default browser to Chrome does work for the Run-REDUCE gui
 > though.
 >
 > Alan
 >
 > It is simply annoying that snap Firefox can't  be used view HTML file URLs
 > unless they are in a sub-directory of your home directory. Iwill affect other
 > software systems that use HTML-based help as well as Reduce.
 >
 > On 01/05/2022 15:36, Francis Wright wrote:
 > > Have you tried making Chrome your default browser?
 > >
 > > Francis
 > >
 > > On 01/05/2022 12:56 pm, Alan Barnes wrote:
 > > > I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now provides
 > > > Firefox via snap rather than a native application.
 > > >
 > > > One effect of this is that it no longer possible to view Reduce html
 > > > documentation via the ReduceCSL Help tab using the default browser:
 > > > Firefox  -- at least if the documentation directory is in a sub-directory
 > > > of a system directory such as /usr/share/reduce,
 > > > /opt/reduce-algebra-code, /usr/local/reduce-algebra-code etc..
 > > >
 > > > I think this is due to snap running in a sandbox with mediated access to
 > > > the underlying Linux system (perhaps the root directory is set to
 > > > something other than /).
 > > >
 > > > There are two possible quick fixes:
 > > >
 > > >       1) set the RedCSL browser command to /usr/bin/google-chrome
 > > >
 > > >       2) move the documentation to a readable sub-directory of /home
 > > > (rather clunky).
 > > >
 > > > However, does anyone know if one can configure the Firefox snap to avoid
 > > > this problem?
 > > >
 > > > Alan Barnes
 > > >
 > > >
 > > >
 > > > _______________________________________________
 > > > Reduce-algebra-developers mailing list
 > > > Red...@li...
 > > > https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers
 >
 >
 > _______________________________________________
 > Reduce-algebra-developers mailing list
 > Red...@li...
 > https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers
 >
 >

   Rainer Schöpf

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Rainer S. <rai...@gm...>]

I suspect that apparmor is blocking the access. There is a special apparmor
profile for snapd.firefox.firefox. Anything in the system logs?

  Rainer


On Sun, 1 May 2022 at 12:56 +0100, Alan Barnes wrote:

 > I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now provides Firefox
 > via snap rather than a native application.
 >
 > One effect of this is that it no longer possible to view Reduce html
 > documentation via the ReduceCSL Help tab using the default browser: Firefox 
 > -- at least if the documentation directory is in a sub-directory of a system
 > directory such as /usr/share/reduce, /opt/reduce-algebra-code,
 > /usr/local/reduce-algebra-code etc..
 >
 > I think this is due to snap running in a sandbox with mediated access to the
 > underlying Linux system (perhaps the root directory is set to something other
 > than /).
 >
 > There are two possible quick fixes:
 >
 >       1) set the RedCSL browser command to /usr/bin/google-chrome
 >
 >       2) move the documentation to a readable sub-directory of /home (rather
 > clunky).
 >
 > However, does anyone know if one can configure the Firefox snap to avoid this
 > problem?
 >
 > Alan Barnes
 >
 >
 >
 > _______________________________________________
 > Reduce-algebra-developers mailing list
 > Red...@li...
 > https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers
 >
 >

   Rainer Schöpf

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Francis W. <f.j...@li...>]

Ah! I only checked on Windows, where CSL REDUCE seems to use the default 
browser. But I'm glad to hear that Run-REDUCE correctly uses the default 
browser.

Francis

On 01/05/2022 5:26 pm, Alan Barnes wrote:
> Simply making Chrome the default browser doesn't affect which browser 
> is used by CSL GUI help. This is Firefox by default.
>
> To change this, one needs to "Select Browser" from the Help menu, 
> check "User" in the dialog that appears and set the command needed to 
> launch the desired browser (google-chrome,  say).
>
> Setting the default browser to Chrome does work for the Run-REDUCE gui 
> though.
>
> Alan
>
> It is simply annoying that snap Firefox can't  be used view HTML file 
> URLs unless they are in a sub-directory of your home directory. Iwill 
> affect other software systems that use HTML-based help as well as Reduce.
>
> On 01/05/2022 15:36, Francis Wright wrote:
>> Have you tried making Chrome your default browser?
>>
>> Francis
>>
>> On 01/05/2022 12:56 pm, Alan Barnes wrote:
>>> I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now 
>>> provides Firefox via snap rather than a native application.
>>>
>>> One effect of this is that it no longer possible to view Reduce html 
>>> documentation via the ReduceCSL Help tab using the default browser: 
>>> Firefox  -- at least if the documentation directory is in a 
>>> sub-directory of a system directory such as /usr/share/reduce, 
>>> /opt/reduce-algebra-code, /usr/local/reduce-algebra-code etc..
>>>
>>> I think this is due to snap running in a sandbox with mediated 
>>> access to the underlying Linux system (perhaps the root directory is 
>>> set to something other than /).
>>>
>>> There are two possible quick fixes:
>>>
>>>       1) set the RedCSL browser command to /usr/bin/google-chrome
>>>
>>>       2) move the documentation to a readable sub-directory of /home 
>>> (rather clunky).
>>>
>>> However, does anyone know if one can configure the Firefox snap to 
>>> avoid this problem?
>>>
>>> Alan Barnes
>>>
>>>
>>>
>>> _______________________________________________
>>> Reduce-algebra-developers mailing list
>>> Red...@li...
>>> https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Francis W. <f.j...@li...>]

Have you tried making Chrome your default browser?

Francis

On 01/05/2022 12:56 pm, Alan Barnes wrote:
> I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now provides 
> Firefox via snap rather than a native application.
>
> One effect of this is that it no longer possible to view Reduce html 
> documentation via the ReduceCSL Help tab using the default browser: 
> Firefox  -- at least if the documentation directory is in a 
> sub-directory of a system directory such as /usr/share/reduce, 
> /opt/reduce-algebra-code, /usr/local/reduce-algebra-code etc..
>
> I think this is due to snap running in a sandbox with mediated access 
> to the underlying Linux system (perhaps the root directory is set to 
> something other than /).
>
> There are two possible quick fixes:
>
>       1) set the RedCSL browser command to /usr/bin/google-chrome
>
>       2) move the documentation to a readable sub-directory of /home 
> (rather clunky).
>
> However, does anyone know if one can configure the Firefox snap to 
> avoid this problem?
>
> Alan Barnes
>
>
>
> _______________________________________________
> Reduce-algebra-developers mailing list
> Red...@li...
> https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers

Re: [Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Alan B. <ala...@gm...>]

Simply making Chrome the default browser doesn't affect which browser is 
used by CSL GUI help. This is Firefox by default.

To change this, one needs to "Select Browser" from the Help menu, check 
"User" in the dialog that appears and set the command needed to launch 
the desired browser (google-chrome,  say).

Setting the default browser to Chrome does work for the Run-REDUCE gui 
though.

Alan

It is simply annoying that snap Firefox can't  be used view HTML file 
URLs unless they are in a sub-directory of your home directory. Iwill 
affect other software systems that use HTML-based help as well as Reduce.

On 01/05/2022 15:36, Francis Wright wrote:
> Have you tried making Chrome your default browser?
>
> Francis
>
> On 01/05/2022 12:56 pm, Alan Barnes wrote:
>> I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now provides 
>> Firefox via snap rather than a native application.
>>
>> One effect of this is that it no longer possible to view Reduce html 
>> documentation via the ReduceCSL Help tab using the default browser: 
>> Firefox  -- at least if the documentation directory is in a 
>> sub-directory of a system directory such as /usr/share/reduce, 
>> /opt/reduce-algebra-code, /usr/local/reduce-algebra-code etc..
>>
>> I think this is due to snap running in a sandbox with mediated access 
>> to the underlying Linux system (perhaps the root directory is set to 
>> something other than /).
>>
>> There are two possible quick fixes:
>>
>>       1) set the RedCSL browser command to /usr/bin/google-chrome
>>
>>       2) move the documentation to a readable sub-directory of /home 
>> (rather clunky).
>>
>> However, does anyone know if one can configure the Firefox snap to 
>> avoid this problem?
>>
>> Alan Barnes
>>
>>
>>
>> _______________________________________________
>> Reduce-algebra-developers mailing list
>> Red...@li...
>> https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers

[Reduce-algebra-developers] Firefox feature on Ubuntu 22.04 (Jammy)

[Alan B. <ala...@gm...>]

I just upgraded Ubuntu from 21.10 (Indri) to Jammy which now provides 
Firefox via snap rather than a native application.

One effect of this is that it no longer possible to view Reduce html 
documentation via the ReduceCSL Help tab using the default browser: 
Firefox  -- at least if the documentation directory is in a 
sub-directory of a system directory such as /usr/share/reduce, 
/opt/reduce-algebra-code, /usr/local/reduce-algebra-code etc..

I think this is due to snap running in a sandbox with mediated access to 
the underlying Linux system (perhaps the root directory is set to 
something other than /).

There are two possible quick fixes:

       1) set the RedCSL browser command to /usr/bin/google-chrome

       2) move the documentation to a readable sub-directory of /home 
(rather clunky).

However, does anyone know if one can configure the Firefox snap to avoid 
this problem?

Alan Barnes

Re: [Reduce-algebra-developers] newfront: tab completion

[Arthur N. <ac...@ca...>]

On Fri, 22 Apr 2022, Andrey G. Grozin wrote:
> Dear Arthur,
> Thank you very much for your work. I confirm that ./rebuild.sh at the 
> top-level directory successfully builds both redpsl and redcsl.

Thank you very much for checking.

> About redfront:
>
> 1. In generic/redfront, make produces
> checking for main in -ltermcap... no
> configure: error: "Please install libtermcap-devel (or similar) first"

As far as I am concerned generic/redfront is there for slightly historical 
reasons, but all questions about it need to be sent to its authors not me.

> 2. In generic/newfront, make cannot find many files. If I do, in generic/,

People are not expected to build in the generic/newfront directory. Those 
files are there so that they get compiled when you do a "make" in a 
top-level directory so as to build redcsl etc

> ln -s redfront/x86_64-unknown-gentoo2.8/libedit/lib .
>
> in generic/newfront, edit Makefile to contain
>
> rfcsl_CPPFLAGS = -DCSL=1 -I. -I../redfront/libedit-20140620-3.1/src
> rfboot_CPPFLAGS = -DBOOT=1 -I. -I../redfront/libedit-20140620-3.1/src
> rfpsl_CPPFLAGS = -DPSL=1 -I. -I../redfront/libedit-20140620-3.1/src
>
Those lines are precisely as they are because it is intended that newfront 
be build in a directory of the form
    trubk/cslbuild/host-triple/redfront

> then make succeeds, and installs rfpsl and rfcsl to bin/. Both of them work, 
> but don't do tab completion for files in in "...
>
I thought than when I had tried I had seen tab completion working when I 
whent 'in "x' where x was the first character of a file that existed. And 
when there were multiple files or a directory to mess with I seems to need 
to type a second tab. But again I do not use tab completion and never have 
so I am not used to what is expected, but all the source code is there for 
you to investigate, and if you patch up redfront so it builds on your 
system by fixing where it tries to look for termcap in ways that are not 
gentoo friendly you can compare the two, which should make any debugging 
easy for somebody who does know what to expect and what to test.

> 3. In cslbuild/x86_64-unknown-gentoo2.8/redfront there are files rfboot, 
> rfcsl, rfpsl. If I copy them to bin/, they work, and do tab completion for 
> files.
>
When you go "make" in the trunk directory that should all happen so you 
should not need to copy things by hand!

> I suppose they are produced from generic/rfront by rebuild.sh, am I right?

The hideous mess of scripts and Makefiles etc tries to do it all 
efforlessly and on Linux veriants that I used regularly by and large they 
do.

> By the way, what is rfboot? At first sight, it does the same as rfcsl.

The CSL version of Reduce is built as follows.
First a version of Reduce ius made and called "bootstrapreduce". That has 
every function making up Reduce compiled into the CSL bytecode format 
which is compact and of acceptable but not great speed. Then 
bootstrapreduce has embedded within itself all the Reduce source code too.
I run profiling to see what functions are most used, and bootstrapreduce 
can then turn those into C++ code which ends up in generated-c/u01.cpp to 
generated-c/u60.cpp. Those are then compiled in with the CSL sources to 
make the "production" version of CSL-Reduce, which in many cases is around 
2.5 times faster than bootstrapreduce. rfboot loads bootstrapreduce for 
the benefit of those who need or want access to it.

> Many thanks,
> Andrey
>
Arthur

Re: [Reduce-algebra-developers] newfront: tab completion

[Andrey G. G. <A.G...@in...>]

Dear Arthur,

Thank you very much for your work. I confirm that ./rebuild.sh at the 
top-level directory successfully builds both redpsl and redcsl.

About redfront:

1. In generic/redfront, make produces

checking for main in -ltermcap... no
configure: error: "Please install libtermcap-devel (or similar) first"

2. In generic/newfront, make cannot find many files. If I do, in generic/,

ln -s redfront/x86_64-unknown-gentoo2.8/libedit/lib .

in generic/newfront, edit Makefile to contain

rfcsl_CPPFLAGS = -DCSL=1 -I. -I../redfront/libedit-20140620-3.1/src
rfboot_CPPFLAGS = -DBOOT=1 -I. -I../redfront/libedit-20140620-3.1/src
rfpsl_CPPFLAGS = -DPSL=1 -I. -I../redfront/libedit-20140620-3.1/src

then make succeeds, and installs rfpsl and rfcsl to bin/. Both of them 
work, but don't do tab completion for files in in "...

3. In cslbuild/x86_64-unknown-gentoo2.8/redfront there are files rfboot, 
rfcsl, rfpsl. If I copy them to bin/, they work, and do tab completion for 
files.

I suppose they are produced from generic/rfront by rebuild.sh, am I right?

By the way, what is rfboot? At first sight, it does the same as rfcsl.

Many thanks,
Andrey

Re: [Reduce-algebra-developers] newfront: tab completion

[Arthur N. <ac...@ca...>]

Two things.
First I HOPE I have checked in small changes to the csl-reduce 
configure.ac scripts so that they can cope with (n)curses on gentoo and 
any other platform where libtinfo needs to be loaded explicitly rather 
than being automatically included just be linking in curses. Please test.

Then when I build and try bin/rfcsl and 'in "some_stuff' then a tab I 
believe that if there are files whose names start with "some_stuff" there 
is completion - though it MAY be necessary to hit tab twice. I have never 
used that but please feel free to inspect the code and propose 
enhancements if you need any. And probably (depending on the age of your 
"old redfront") the files in generic/redfront rather than generic/newfront 
woulkd give you an older version to compare and maybe it could be that 
building that for yourself would take you back in time in case that 
behaves better. But please if you do that work out what changes in 
required in the newer version and report them in form that is carefully 
portable across as many operating systems as you can!

Arthur

On Thu, 21 Apr 2022, Andrey G. Grozin wrote:

> In the old redfront, if I write
>
> in "abc
>
> and press tab, redfront does tab completion over files in the current 
> directory. Very convenient.
>
> newfront does not do tab completion. Is it a feature or a bug?
>
> Andrey
>
>
> _______________________________________________
> Reduce-algebra-developers mailing list
> Red...@li...
> https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.sourceforge.net%2Flists%2Flistinfo%2Freduce-algebra-developers&data=05%7C01%7Cacn1%40universityofcambridgecloud.onmicrosoft.com%7C533ea121dcb84e184ba008da2344dbd2%7C49a50445bdfa4b79ade3547b4f3986e9%7C0%7C0%7C637861075900178666%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=l0IZp%2Bjs6IWXoCmin7x7iq%2BFJ8ButeFZ2UJJiPhKtA0%3D&reserved=0
>

Re: [Reduce-algebra-developers] reduce-mode.el: bugs in syntax highlighting

[Francis W. <f.j...@li...>]

Thanks for the report. I appear to have fixed most of these bugs in my 
development version. I'll try to fix the remaining problem as soon as I 
have some time and release an update. The remaining problem is that "y 
such that" is blue and I think that's because my font-lock code is 
seeing this as a function composition.

Francis

On 21/04/2022 4:09 am, Andrey G. Grozin wrote:
> In emacs, in a reduce file, try
>
> x(1):=1$
> x(1) :=1$
> x(1):= 1$
> x(1) := 1$
>
> In the first 2 cases, =, 1, and $ are green. Why?
> In the last 2 cases, = is green. Why?
>
> for all x,y let f(x,y)=0;
> for all x,y clear f(x,y);
> for all x,y such that x<0 let f(x,y)=0;
> for all x,y such that x<0 clear f(x,y);
>
> In the 2-nd line, y and clear are blue. Why?
> In the 3-rd line, y, such, and that are blue. Why?
> In the 4-th line, y, such, that, and clear are blue. Why?
> Wouldn't it be natural to always have clear in magenta, like let?
>
> Andrey
>
>
> _______________________________________________
> Reduce-algebra-developers mailing list
> Red...@li...
> https://lists.sourceforge.net/lists/listinfo/reduce-algebra-developers

[Reduce-algebra-developers] newfront: tab completion

[Andrey G. G. <A.G...@in...>]

In the old redfront, if I write

in "abc

and press tab, redfront does tab completion over files in the current 
directory. Very convenient.

newfront does not do tab completion. Is it a feature or a bug?

Andrey

[Reduce-algebra-developers] reduce-mode.el: bugs in syntax highlighting

[Andrey G. G. <A.G...@in...>]

In emacs, in a reduce file, try

x(1):=1$
x(1) :=1$
x(1):= 1$
x(1) := 1$

In the first 2 cases, =, 1, and $ are green. Why?
In the last 2 cases, = is green. Why?

for all x,y let f(x,y)=0;
for all x,y clear f(x,y);
for all x,y such that x<0 let f(x,y)=0;
for all x,y such that x<0 clear f(x,y);

In the 2-nd line, y and clear are blue. Why?
In the 3-rd line, y, such, and that are blue. Why?
In the 4-th line, y, such, that, and clear are blue. Why?
Wouldn't it be natural to always have clear in magenta, like let?

Andrey

[Reduce-algebra-developers] The complete victory [was: A partial victory]

[Andrey G. G. <A.G...@in...>]

In generic, I did

ln -s redfront/x86_64-unknown-gentoo2.8/libedit/lib .

I edited gereric/newfront/Makefile to contain

rfcsl_CPPFLAGS = -DCSL=1 -I. -I../libedit/src
rfboot_CPPFLAGS = -DBOOT=1 -I. -I../libedit/src
rfpsl_CPPFLAGS = -DPSL=1 -I. -I../libedit/src

After that, make succeeds, and installs rfpsl, rfcsl, rfboot in bin. So, 
my favorite way to use reduce, rfpsl, now works!

What is rfboot?

Andrey

Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Arthur C. N. <ac...@ca...>]

If you are confident pkgconfig is installed and set up. And still that would be used in configure.ac along with other curses etc detection stuff.

Get Outlook for Android<https://aka.ms/AAb9ysg>
________________________________
From: Andrey G. Grozin <A.G...@in...>
Sent: Wednesday, April 13, 2022 10:02:22 AM
To: Arthur Charles Norman <ac...@ca...>
Cc: red...@li... <red...@li...>
Subject: Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

On Wed, 13 Apr 2022, Andrey G. Grozin wrote:
> By the way, I thought that the canonical way to find $LIBS for ncurses is
>
> grozin@bilbo ~ $ ncurses6-config --libs
> -L/usr/lib64 -lncurses -ltinfo
Wrong! the *really* correct and modern way is

grozin@bilbo ~ $ pkgconf ncurses --libs
-lncurses -ltinfo

Andrey

Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Andrey G. G. <A.G...@in...>]

On Wed, 13 Apr 2022, Andrey G. Grozin wrote:
> By the way, I thought that the canonical way to find $LIBS for ncurses is
>
> grozin@bilbo ~ $ ncurses6-config --libs
> -L/usr/lib64 -lncurses -ltinfo
Wrong! the *really* correct and modern way is

grozin@bilbo ~ $ pkgconf ncurses --libs
-lncurses -ltinfo

Andrey

Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Andrey G. G. <A.G...@in...>]

On Wed, 13 Apr 2022, Arthur Charles Norman wrote:
> It is really
> unlikely that Gentoo breaks it in any deep way, but lat time I tried to install a Gentoo it was painfully time consuming . I may
> have to spend that time, but maybe next week (is not just now) you can tell me just what fails and I may then be able to spot or
> suggest fixes.
In my experience, installing Gentoo requires 1 full day of hard work. 
Don't spend your time doing it. I'll better create a user for you on my 
notebook, so that you will be able to log in via ssh. We'll need to 
agree upon some specific times, because this notebook is usually 
switched off during nights (Novosibirsk time); of course, I can leave it 
switched on when needed.

Gentoo is an excellent development platform: it has all tools and 
libraries of (usually) latest versions. True, it is somewhat too flexible: 
2 different Gentoo installations can differ very much.

By the way, I thought that the canonical way to find $LIBS for ncurses is

grozin@bilbo ~ $ ncurses6-config --libs
-L/usr/lib64 -lncurses -ltinfo

Andrey

Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Arthur C. N. <ac...@ca...>]

When I am back from what was intended as a break some simple changes in configure.ac where nurses is detected can add -ltinfo

Get Outlook for Android<https://aka.ms/AAb9ysg>
________________________________
From: Andrey G. Grozin <A.G...@in...>
Sent: Wednesday, April 13, 2022 6:03:10 AM
To: Arthur Charles Norman <ac...@ca...>
Cc: red...@li... <red...@li...>
Subject: A partial victory [was: trying to build the current reduce (r6281)]

I've managed to build csl reduce. In an absolutely barbaric method, no
automation. I'll document it here for the benefit of those who has a
separate libtinfo.

Toplevel directory:
./configure --with-csl
make csl
Linking of csl fails. In cslbuild/x86_64-unknown-gentoo2.8/csl I paste the
failed linking command to the command line, append -ltinfo and execute. It
succeeds. Toplevel directory:
make csl
Linking of bootstrapreduce fails. Again go to
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute, go to the
toplevel directory.
make csl
Linking of reduce fails. Again go to
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute. Then,
still in this directory,
make reduce.img
make csl.img

bin/redcsl, bin/redcsl -w, bin/csl work fine. Still haven't figured out
how to build redfront. And what is newfront? Should I build it instead of
redfront?

Andrey

Re: [Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Arthur C. N. <ac...@ca...>]

When I am back from what was intended to be a break I can put simple and almost obvious changes in configure.ac to link in libtinfo on systems where nurses does not do that for you.
Newfront is redfront but with portability fixes so it could be built on all the platforms we usually work with. It is really unlikely that Gentoo breaks it in any deep way, but lat time I tried to install a Gentoo it was painfully time consuming . I may have to spend that time, but maybe next week (is not just now) you can tell me just what fails and I may then be able to spot or suggest fixes. Arthur
Just back from looking for nightjars and finding elephant.

Get Outlook for Android<https://aka.ms/AAb9ysg>
________________________________
From: Andrey G. Grozin <A.G...@in...>
Sent: Wednesday, April 13, 2022 6:03:10 AM
To: Arthur Charles Norman <ac...@ca...>
Cc: red...@li... <red...@li...>
Subject: A partial victory [was: trying to build the current reduce (r6281)]

I've managed to build csl reduce. In an absolutely barbaric method, no
automation. I'll document it here for the benefit of those who has a
separate libtinfo.

Toplevel directory:
./configure --with-csl
make csl
Linking of csl fails. In cslbuild/x86_64-unknown-gentoo2.8/csl I paste the
failed linking command to the command line, append -ltinfo and execute. It
succeeds. Toplevel directory:
make csl
Linking of bootstrapreduce fails. Again go to
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute, go to the
toplevel directory.
make csl
Linking of reduce fails. Again go to
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute. Then,
still in this directory,
make reduce.img
make csl.img

bin/redcsl, bin/redcsl -w, bin/csl work fine. Still haven't figured out
how to build redfront. And what is newfront? Should I build it instead of
redfront?

Andrey

[Reduce-algebra-developers] A partial victory [was: trying to build the current reduce (r6281)]

[Andrey G. G. <A.G...@in...>]

I've managed to build csl reduce. In an absolutely barbaric method, no 
automation. I'll document it here for the benefit of those who has a 
separate libtinfo.

Toplevel directory:
./configure --with-csl
make csl
Linking of csl fails. In cslbuild/x86_64-unknown-gentoo2.8/csl I paste the 
failed linking command to the command line, append -ltinfo and execute. It 
succeeds. Toplevel directory:
make csl
Linking of bootstrapreduce fails. Again go to 
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute, go to the 
toplevel directory.
make csl
Linking of reduce fails. Again go to
cslbuild/x86_64-unknown-gentoo2.8/csl, append -ltinfo, execute. Then, 
still in this directory,
make reduce.img
make csl.img

bin/redcsl, bin/redcsl -w, bin/csl work fine. Still haven't figured out 
how to build redfront. And what is newfront? Should I build it instead of 
redfront?

Andrey