Package libpcre as a standard package

[charpent]

Rationale : R (standard package) requires libpcre (in a non-default configuration), but stopped to package it with version 3.3.0.

See this ​discussion on sage-devel, but look also ​here and contribute ​here.

Tarball : ​https://ftp.pcre.org/pub/pcre/pcre-8.40.tar.gz

[charpent]

New commits:

​a8d77b5

Initial packaging of pcre : unconditionnal installation.

[charpent]

The initial packaging builds without problem and passes its own testsuite ; the resulting Sage passes ptestlong with only the usual transient error on simplicial_complex.py (which passes standalone).

==> needs_review.

[jdemeyer]

In the light of #20692, please remove this patch block (especially given that there are no patches):

# Patch sources.
for patch in ../patches/*.patch; do
    [ -r "$patch" ] || continue  # Skip non-existing or non-readable patches
    patch -p1 <"$patch"
    if [ $? -ne 0 ]; then
        echo >&2 "Error applying '$patch'"
        exit 1
    fi
done

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​21ce3e7

Removed patches handling.

[charpent]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​21ce3e7

Removed patches handling.

Builds OK, passes its own test suite.

[charpent]

Sage-7.6.beta4 + #20523 + #21767 + #21770 (this patch) passes ptestlong witht two transient errors :

----------------------------------------------------------------------
sage -t --long src/sage/homology/simplicial_set_morphism.py  # Bad exit: 2
sage -t --long src/sage/homology/simplicial_complex.py  # 1 doctest failed
----------------------------------------------------------------------

Both tests pass when run standalone.

This should close the R saga (for now) by allowing the current interfaces to work with R >= 3.3.x.

Still needs_review

[charpent]

Merged with #20523

This ticket should be marked as invalid by someeone with the right rights...

[charpent]

After discussion of (potential) #20523 reviewers, I now think that this ticket should be kept, reviewed and validated. Pleas DON'T invalidate it.

[jdemeyer]

[pcre-8.39] WARNING: spkg-install is not executable, making it executable

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​bfc26ef

Wups. chmod +x spkg-install...

[charpent]

Replying to jdemeyer:

[pcre-8.39] WARNING: spkg-install is not executable, making it executable

Done. needs_review

[dimpase]

I wonder if pcre also pops up in polymake-related tickets (which do need a lot of perl)...

[charpent]

Updated the address of the original tarball (patchbots seem to be unable to use an FTP address...).

[jdemeyer]

I would suggest to make the package optional for now and only make it standard when really needed.

[charpent]

Replying to jdemeyer:

I would suggest to make the package optional for now and only make it standard when really needed.

We've already been there (see previous saga...) :

• R is a standard package
• R depends on PCRE starting 3.3.x
• A standard package cannot depend on an optional package

Therefore pcre has to be standard. QED.

[charpent]

And, yes, it needs_review.

[jdemeyer]

Replying to charpent:

• R version 3.2.4-revised is a standard package
• R version 3.3.x depends on PCRE
• A standard package cannot depend on an optional package

I stand by my proposal: make it standard when really needed.

[charpent]

Replying to jdemeyer:

Replying to charpent:

• R version 3.2.4-revised is a standard package

And is about 1 year late on current R. Having an up-to-date R version is a sine qua non to get help on R-help, and to be able to install recent versions of the >10 000 R packages. Which are a not inconsiderable resource to those lowly life-forms called applied statisticians...

• R version 3.3.x depends on PCRE
• A standard package cannot depend on an optional package

I stand by my proposal: make it standard when really needed.

IMHO it *is* really needed. But it happens that I need R for my everyday work. Do you ?

[jdemeyer]

Replying to charpent:

Replying to jdemeyer:

Replying to charpent:

• R version 3.2.4-revised is a standard package

And is about 1 year late on current R. Having an up-to-date R version is a sine qua non to get help on R-help, and to be able to install recent versions of the >10 000 R packages. Which are a not inconsiderable resource to those lowly life-forms called applied statisticians...

That may be all true, but I don't see how that matters here.

• R version 3.3.x depends on PCRE
• A standard package cannot depend on an optional package

I stand by my proposal: make it standard when really needed.

IMHO it *is* really needed.

...when we upgrade to R-3.3.0, which is not the case yet.

[charpent]

Replying to jdemeyer:

Replying to charpent:

Replying to jdemeyer:

Replying to charpent:

• R version 3.2.4-revised is a standard package

And is about 1 year late on current R. Having an up-to-date R version is a sine qua non to get help on R-help, and to be able to install recent versions of the >10 000 R packages. Which are a not inconsiderable resource to those lowly life-forms called applied statisticians...

That may be all true, but I don't see how that matters here.

• R version 3.3.x depends on PCRE
• A standard package cannot depend on an optional package

I stand by my proposal: make it standard when really needed.

IMHO it *is* really needed.

...when we upgrade to R-3.3.0, which is not the case yet.

#20523 is waiting on this ticket for review (and has been for a looong time : it has been delayed by the openssl brouhaha).

[jdemeyer]

Replying to charpent:

#20523 is waiting on this ticket for review (and has been for a looong time : it has been delayed by the openssl brouhaha).

Then make it optional here and standard in #20523.

[charpent]

Replying to jdemeyer:

Replying to charpent:

#20523 is waiting on this ticket for review (and has been for a looong time : it has been delayed by the openssl brouhaha).

Then make it optional here and standard in #20523.

I can't see the point... Care to explain ?

[jdemeyer]

Replying to charpent:

I can't see the point... Care to explain ?

I don't see the point of having a standard package which is not used by anything in Sage.

[charpent]

Replying to jdemeyer:

Replying to charpent:

I can't see the point... Care to explain ?

I don't see the point of having a standard package which is not used by anything in Sage.

OK. If I follow you :

• PCRE is not used by anything in Sage
• PCRE is used by R

Therefore, by modus tollens:

• R is not in Sage.

We should submit this interesting conclusion to sage-devel : the discussion would probably be lively and entertaining (for questionable values of "entertaining"...).

[jpflori]

@emmanuel: can you update to 8.40 pls?

[charpent]

Replying to jpflori:

@emmanuel: can you update to 8.40 pls?

I'm on it. Count about 15 minutes to post an update patch, about 2-3 hours to compile and ptestlong it.

BTW : since #21767 and the present patch have been merged in #20523, I'll have to update it also before review.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ef48f71

Upgraded to pcre 8.40

[charpent]

Upgrade as requested. Therefore, tarball modified accordingly.

---

New commits:

​ef48f71

Upgraded to pcre 8.40

[charpent]

• Passes its own testsuite.
• (#21567 + #21770 + #20523 remerged with the two previous ones) passes ptestlong with two failures :

  ----------------------------------------------------------------------
  sage -t --long src/sage/homology/simplicial_complex.py  # 1 doctest failed
  sage -t --long src/sage/rings/polynomial/multi_polynomial_ideal.py  # 1 doctest failed
  ----------------------------------------------------------------------
  

The first one is transient (passes when ran standalone) ; the second one is a well-documented, ticketed and not yet fixed difference on the expression of the expected result.

==> needs_review for good.

[jpflori]

Looks ok.

[jdemeyer]

Is the --enable-jit flag really required? It already caused problems on Cygwin and it is breaking on Solaris too (#24628).