re2c 1.1 error

[danielluke (Daniel J. Luke)]

I use re2c with spamassassin, and with 1.1, sa-compile says:

libc++abi.dylib: terminating with uncaught exception of type std::out_of_range: basic_string
/opt/perlbrew/perls/perl-5.28.0/bin/sa-compile requires re2c for proper operation.

This is probably an upstream bug, I haven't reported it there (yet). We may want to consider reverting to version 1.0.3 which works.

[danielluke (Daniel J. Luke)]

Upstream bug: ​https://github.com/skvadrik/re2c/issues/214

[pmetzger (Perry E. Metzger)]

It might be that the abi changed a bit, what happens if you force a recompilation of spam assassin? If that's really the problem then we'll need to rev bump things that use re2c.

[pmetzger (Perry E. Metzger)]

Actually, what's the "/opt/perlbrew" business? This isn't a macports version of spam assassin you're running?

[pmetzger (Perry E. Metzger)]

Okay, I re-familiarized myself with the situation. re2c installs a single executable which is used to compile things, not any libraries. The new re2c passes something like 1400 tests in the internal test suite. I tried it on my own programs that use re2c and it is working. My guess is something weird is happening. Can you try to figure out why it is that your install of sa-compile (which doesn't look to be from macports?) is not finding the binary, or why it is otherwise crashing?

[danielluke (Daniel J. Luke)]

Replying to pmetzger:

It might be that the abi changed a bit

the re2c port provides a binary and a manpage, so ABI isn't relevant.

[danielluke (Daniel J. Luke)]

Replying to pmetzger:

Can you try to figure out why it is that your install of sa-compile (which doesn't look to be from macports?) is not finding the binary,

It is finding the binary.

or why it is otherwise crashing?

it is crashing when sa-compile runs 're2c -V'

[danielluke (Daniel J. Luke)]

commenting out the check in sa-compile allows it to complete successfully

[pmetzger (Perry E. Metzger)]

Okay, definitively an upstream error. Lets see if we can't patch this, it seems kind of ridiculous for the -V flag to dump core.

[danielluke (Daniel J. Luke)]

Replying to pmetzger:

Actually, what's the "/opt/perlbrew" business?

This isn't really relevant from the bug report - but I gave up on using MacPorts perl for anything other than MacPorts dependencies a long time ago. Our project's insistence of trying to support modules with every perl version (and then not updating which perl is the 'default' perl for long periods of time) + our inability to track the 'current' versions of most of the p5 ports means it doesn't make sense for me to use it for any 'real' work.

We could fix 90% of the problem of trying to use MacPorts perl by just having one perl5 port that installs the latest perl5 and doing something to revbump all the p5 ports whenever we update it to a new perl5 minor release. We could fix 100% of the problem if we developed something to integrate with cpan(m) -- but there's been little to no interest in those solutions and I haven't had time to contribute anything real to help in this area back to the project.

Using perlbrew to mange that stuff works much better for my perl needs.

[pmetzger (Perry E. Metzger)]

It ceased to be relevant when it became obvious what the problem was.

Anyway, lets see if we can't track down what's causing the core dump. The re2c maintainer has been very responsive in the past with me.

[pmetzger (Perry E. Metzger)]

BTW, I think I probably am de facto the owner of this ticket, since my update did this.

[pmetzger (Perry E. Metzger)]

In cba5769fbc1db6305e243880474667f82a734e3a/macports-ports (master):

re2c: fix a crash when running "re2c -V"

closes: #57063