Latte binary count called latte-count on archlinux

[videlec]

In my setup (see below for details), the program count that sage expects to be from LattE (integral points count in polytope) is actually a binary from llvm.

Steps To Reproduce

With latte installed in SageMath 10.2 in a conda environment, the following computation fails

sage: V = [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
....:      [0, 0, 0, 1, 0, 0, 0, 0, 0, 1],
....:      [0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
....:      [0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
....:      [0, 0, 0, 1, 0, 0, 1, 0, 0, 0],
....:      [0, 0, 0, 1, 0, 1, 0, 0, 0, 0],
....:      [0, 0, 0, 1, 1, 0, 0, 0, 0, 0],
....:      [0, 0, 1, 0, 0, 0, 1, 0, 0, 0],
....:      [0, 1, 0, 0, 0, 1, 0, 0, 0, 0],
....:      [1, 0, 0, 0, 1, 0, 0, 0, 0, 0]]
sage: P = Polyhedron(V)
sage: P.integral_points_count()

Expected Behavior

Same as

sage: len(P.integral_points())
10

Actual Behavior

File ~/sage/geometry/polyhedron/base_QQ.py:214, in Polyhedron_QQ.integral_points_count(self, verbose, use_Hrepresentation, explicit_enumeration_threshold, preprocess, **kwds)
    211     use_Hrepresentation = True
    213 from sage.interfaces.latte import count
--> 214 return count(
    215         p.cdd_Hrepresentation() if use_Hrepresentation else p.cdd_Vrepresentation(),
    216         cdd=True,
    217         verbose=verbose,
    218         **kwds)

File ~/sage/interfaces/latte.py:175, in count(arg, ehrhart_polynomial, multivariate_generating_function, raw_output, verbose, **kwds)
    173     else:
    174         err = ":\n" + err
--> 175     raise RuntimeError("LattE integrale program failed (exit code {})".format(ret_code) + err.strip())
    177 ans = bytes_to_str(ans)
    179 # There's an error handler below that uses the numOfLatticePoints
    180 # file created by latte, so we can't cleanup() the temporary
    181 # directory here. Instead we have to clean it up before the
    182 # (several) return statements.

RuntimeError: LattE integrale program failed (exit code 2):
usage: /usr/bin/count <expected line count>

Additional Information

No response

Environment

- **OS**: archlinux with sage in a conda environment, LattE not installed in conda and llvm installed on the system
- **Sage Version**: 10.2

Checklist

• I have searched the existing issues for a bug report that matches the one I want to file, without success.
• I have read the documentation and troubleshoot guide

[maxale]

It works fine in https://sagecell.sagemath.org also running SageMath version 10.2

[videlec]

It works fine in https://sagecell.sagemath.org also running SageMath version 10.2

Thanks for trying this out! Do you know which version of LattE runs on sagecell?

[videlec]

The problem comes from the fact that my /usr/bin/count is not from the LattE suite but llvm. The LattE binaries are called: latte-count, latte-maximize, latte-minimize (see details on the package page). Not sure to know what solution should be adapted in sage...

[maxale]

@videlec: Nice find! It's worth it to update the title of this issue accordingly.

[videlec]

@antonio-rojas @mkoeppe any idea how to solve this consistently?

[antonio-rojas]

@antonio-rojas @mkoeppe any idea how to solve this consistently?

There are two issues here: first one is Sage accepting any executable named count. This could be easily solved by adding an is_functional method to the Latte_count class in src/sage/features/latte.py that checks whether it actually comes from latte (by testing eg. that the output of the count command contains the "LattE" string).

The second issue is the conda-installed Sage choosing the system count executable over the conda one. I know nothing about conda, but I would expect stuff in conda to be run with a PATH that prioritizes conda's own bin paths over the system ones. Is that not the case?

[videlec]

@antonio-rojas @mkoeppe any idea how to solve this consistently?

There are two issues here: first one is Sage accepting any executable named count. This could be easily solved by adding an is_functional method to the Latte_count class in src/sage/features/latte.py that checks whether it actually comes from latte (by testing eg. that the output of the count command contains the "LattE" string).

This I can easily do. Thanks for the suggestion.

The second issue is the conda-installed Sage choosing the system count executable over the conda one. I know nothing about conda, but I would expect stuff in conda to be run with a PATH that prioritizes conda's own bin paths over the system ones. Is that not the case?

There is a priority of conda path over system paths. However, in my setup, LattE is not installed in conda.

[antonio-rojas]

The second issue is the conda-installed Sage choosing the system count executable over the conda one. I know nothing about conda, but I would expect stuff in conda to be run with a PATH that prioritizes conda's own bin paths over the system ones. Is that not the case?

There is a priority of conda path over system paths. However, in my setup, LattE is not installed in conda.

OK, that makes sense then. I had interpreted "With latte installed in SageMath 10.2 in a conda environment" as you had installed latte from within sage.

[videlec]

#38678 does implement is_functional.

However, it could have worked in the initial setup (sage from conda, LattE from the system) if there was a way for sage to "discover automagically" count versus latte-count.

[antonio-rojas]

However, it could have worked in the initial setup (sage from conda, LattE from the system) if there was a way for sage to "discover automagically" count versus latte-count.

As much as I would personally benefit from it (as it would allow me to get rid of the only remaining permanent downstream patch), I really don't think it is sage's job to cater to all possible downstream variations in the naming of executables. That could easily become a mess if each distro does its own thing.

The reasonable solution would be to convince upstream to rename its executables to something less generic.