svn commit: r328240 - in head: etc/mtree lib lib/libc/regex lib/libc/tests/regex lib/libregex lib/libregex/tests share/mk

[Kyle Evans]

On Mon, Jan 22, 2018 at 4:26 AM, Konstantin Belousov
<kostikbel at gmail.com> wrote:
> On Mon, Jan 22, 2018 at 02:44:41AM +0000, Kyle Evans wrote:
>> Author: kevans
>> Date: Mon Jan 22 02:44:41 2018
>> New Revision: 328240
>> URL: https://svnweb.freebsd.org/changeset/base/328240
>>
>> Log:
>>   Add libregex, connect it to the build
>>
>>   libregex is a regex(3) implementation intended to feature GNU extensions and
>>   any other non-POSIX compliant extensions that are deemed worthy.
>>
>>   These extensions are separated out into a separate library for the sake of
>>   not cluttering up libc further with them as well as not deteriorating the
>>   speed (or lack thereof) of the libc implementation.
>>
>>   libregex is implemented as a build of the libc implementation with LIBREGEX
>>   defined to distinguish this from a libc build. The reasons for
>>   implementation like this are two-fold:
>>
>>   1.) Maintenance- This reduces the overhead induced by adding yet another
>>   regex implementation to base.
>>
>>   2.) Ease of use- Flipping on GNU extensions will be as simple as linking
>>   against libregex, and POSIX-compliant compilations can be guaranteed with a
>>   REG_POSIX cflag that should be ignored by libc/regex and disables extensions
>>   in libregex. It is also easier to keep REG_POSIX sane and POSIX pure when
>>   implemented in this fashion.
> You are doing very fragile and unmaintainable trick on all consumers
> there.  Your libregex.so exports the same symbols under the same version
> as the libc does. In other words, we now provide two binary-incompatible
> callable symbols, and selection of the symbol by the consumer depends on
> the DT_NEEDED order and interposing.  For instance, if some program loads
> a module linked to your libregex, the program behaviour suddenly changes.
>
> Since the library provides incompatible implementation, it must use
> different versions for the symbols, at least to save others time to
> debug the mess.

What's the best way that you see, going forward?

I'm inclined to throw a Symbol.map into libregex using FBSD_1.1...
these interfaces are otherwise stable stable within the two respective
libraries, so I don't see that causing too much pain in the future
because symbol version changes should be rare.

On the other hand, I could see wanting to use something more like
FBSD_LIBREGEX_1.0 so that if the situation does come up one doesn't
need to double-check that they're not colliding with the other
implementation.