On Thu, Feb 19, 2026 at 07:08:52PM +0100, Patrice Dumas wrote: > On Thu, Feb 19, 2026 at 05:53:15PM +0000, Gavin Smith wrote: > > > The problem is not a grave one, in any case. The programs and > > > libraries do build and pass all the tests. A complication warning is > > > not a problem, and I can fix it by adding an explicit prototype in > > > that source file. So the solution of this could be deferred to the > > > next release, if you are okay with that. > > > > Chances are if this code was actually run, there would be the "Free > > to wrong pool" error (on MS-Windows), depending on where this strndup > > function is actually coming from. If it's a version defined by gnulib, > > it would not be compatible with the 'free' function in the file as this > > is Perl's version of 'free'. > > There is a test using this function, so if the gnulib function was > actually used, there would be the "Free to wrong pool" error triggered.
Thanks for fixing my broken commit. After making the change, I ran "make" under the tta/C subdirectory and saw a successful build, with no error messages. The tests all passed, as well. What I hadn't realised was that I had configured with --disable-perl-xs (for testing). So some, but not all, of the code under tta/C was rebuilt, but not the code I had just edited. The test suite was running the non-XS code, so again not the code I had edited. I'm not sure how I could avoid such mistakes, other than by experience. It's one of the downsides of having alternative implementations that are used in different configurations.
