Joe Orton <[EMAIL PROTECTED]> writes: [...]
> The ltmaj fix is not necessary for APR, since APR embeds the major > version in the library name to allow parallel installs of different > major versions, so always passing CURRENT == AGE as the libtool > interface versions is OK. > > I'm a little hesitant about gratuitously changing the soname just > before a major release, and after 1.0 it will be too late. No worries, it's a minor polish issue that allows the APR_DOTTED_VERSION to match the generated soname. I brought it up now because it wasn't an issue in the 0.9.X releases. If binary compatibility has been stable since apr started generating libapr-1.so.0.0.0, then changing the soname would be a bad idea. OTOH if it has broken recently, then bumping the number might be worth consideration. > Is there a reason why you need the get-version.sh script included in > APR to behave like this, rather than just using a local modification > in the copy in the libapreq2 tree? No- libapreq2's dependency on apr is solely through apr-config. Whatever decision is made on the soname will not impact libapreq2. However, adjusting its name to apr-1-config will likely us avoid problems in the future (since we're tied to httpd-2.0's ABI). -- Joe Schaefer
