DBI revived

[Tux]

DBI 1.644 released

This is more news than just a new version.

On 2024-08-11 I asked Tim Bunce if he would be able to do a new release, as the last one was over 4 years ago, and there have been several fixes since, amongst which a CVE fix and a followup CVE fix. Those were the reason I asked.

What happened then was unexpected and more than I bargained for: he said it was extrememly unlikely he'd be involved anymore. He said he trusted me and handen me the ownership of DBI.

I then asked for help and got it. There is more to do, so lets join knowledge and not wait another 4 years.

PLEASE test and test and test and test, with all the database drivers you are able to use on all Operating systems you can work on with as many weird combinations as possible.

I tested on every released perl since 5.8.0 threaded and non-threaded on Linux, and followed the dev-releases on the cpantesters (which surfaced a Windows bug).

As this module is the basis for every DBD in the perl toolchain, it is quite important we keep it going strong.

It feels like an honor to be in this position, and I sincerely hope I can hold the fortress.

irc.perl.org:6667/#dbi is our main communication channel

[cavac]

I'm not sure if "Congratulations" are in order here (DBI is a rather critical module and an awful lot of work), but "Thank you for your service" certainly is.

Couple of small details regarding the 1.644 POD: Please mention youself as the current author/maintainer in the "Author" section. And, secondly, as far as i can tell Tim Bunce's webpage has moved to a different domain, this probably needs an update in the POD.

[Discipulus]

..what to say.. many thanks Tux!

to take care of this critical piece of our ecosystem.. 2059 total dependants!!

I Have succesfully tested it on my strawberry portable distros: perl5.12-32bit perl5.22.64bit perl5.24.64bit perl5.26.64bit perl-5.26.64bit-PDL perl5.28.32bit perl5.28-64bit perl5.32.64bit

As I have tried to help you I'm curios about the windows bug you spotted: if you can easily point me to quick look of the code you changed I'd like it ( update: here, thanks!)

L*

[etj]

Something that will help, though I can see will be a certain amount of effort (I remember converting PDL from Travis to GitHub Actions (GHA)), will be to use GHA for DBI, better.

The current CI config is somewhat minimal: many Perl versions, but that's all: https://github.com/perl5-dbi/dbi/blob/master/.github/workflows/test.yml

A good example that has lots of weird and wonderful DBDs is DBIx::Class's, but that's only on Travis, so would need updating to GHA; the main Travis config is at https://github.com/Perl5/DBIx-Class/blob/maint/0.0828xx/.travis.yml, which calls various scripts in https://github.com/Perl5/DBIx-Class/tree/maint/0.0828xx/maint/travis-ci_scripts.

[DSolimano]

Thank you! I've tested new DBI on my Cygwin and Windows smokers such as they are, all successful.

I'll see if I can test some of the DBDs and take a peek at those failure reports, though as someone else mentioned I have seen files being held open on Cygwin in one or two other modules and had to submit workarounds, so that is a likely cause.

[Danny]

Hi Tux. Regarding your request in CB for testing on Cygwin, I would be happy to do tests with Cygwin. Did you have specific tests in mind or just want to know if 1.644 works?

[Tux]

Thanks for stepping up: Ideally I'd like someone to test DBI from git checkouts with as many supported DBDs as possible.

Any subset of that is valuable.

I was able to get Cygwin up and running on a Windows system I had access to, but I was not able to reproduce the failures reported on here for Cygwin. (Yes, Ik know, those are fails on Text::CSV_XS, but that is what is used for DBD::CSV, so related)

[Corion]

binmode() on closed filehandle $fh at t/85_util.t line 286.
print() on closed filehandle $fh at t/85_util.t line 288.
print() on closed filehandle $fh at t/85_util.t line 289.
binmode() on closed filehandle $fh at t/85_util.t line 306.

I would assume that for some dark reason, the open my $fh, ">", $fnm; in line 285 fails. Most likely, because the file is still opened by another process, be it either a parallel test run or a virus scanner or something like that.

To test this theory, consider adding a ... or die "Couldn't overwrite '$fnm': $!/$^E";. If that really turns out to be the case, doing a loop for (say) 10 times to try to create the file might be a workaround, or generating a filename using File::Templ instead of using the hardcoded filename.

[Danny]

Text::CSV_XS from github passed all tests on my cygwin. The main cygwin distribution recently updated to Perl 5.40.0 and DBI 1.644. The DBI 1.644 from github also passed all tests on my machine, but a number of them were skipped.

For testing DBDs is the idea to go down the list and run 'make test' on each one?

[Danny]

It seems all those fails are on x86_64-cygwin-thread-multi-quadmath except that the Perl 5.41.1 also has one on x86_64-cygwin-thread-multi-ld. I'm on x86_64-cygwin-threads-multi, but I'll test it anyway.

[1nickt]

Hi Tux

Successfully installed on Mac OS with Perl 5.38.2. Report sent to cpan-testers. Also ran our internal test suite of 5326 tests; no failures. Thanks for your service!