Nicolas, Roland can state the exact reason for the design choice but as far as I know it was an explicit choice *not* to use interposers. There are a couple of warnings in his notes to verify for each update that no interposing *must* happen.
Olga On Fri, Apr 23, 2010 at 1:41 AM, Nicolas Williams <nicolas.willi...@oracle.com> wrote: > On Thu, Apr 22, 2010 at 05:32:09PM -0600, bugmail-sen...@sun.com wrote: >> *Synopsis*: mail delivery hangs due to ksh93, via AST, using an old and slow >> getcwd() algorithm >> >> CR 6946515 changed on Apr 22 2010 by <User 1-5Q-101> >> >> === Field ============ === New Value ============= === Old Value >> ============= >> >> Evaluation New Note >> ====================== =========================== >> =========================== > > The evaluation says that libc's getcwd() uses malloc (when the pathname > argument is NULL), and since AST's malloc() does not actually interpose > on libc's malloc(), getcwd() will end up using libc's malloc, which > creates problems (the note refers to AST's malloc being async-signal- > safe while libc's isn't). > > That's not a big deal since _ast_getcwd() can just allocate its own > buffer and then call the real getcwd() with that. > > However, this brings up a different issue: why are the AST libc > replacements not interposers? Why are are the AST libc replacements > simply CPP renames of the function names? > > It's generally dangerous to have more than one allocator in use in a > process. For ksh93, being so tightly coupled with libast, that's not an > issue, but I suspect eventually this will bite someone. > > Nico > -- > _______________________________________________ > ksh93-integration-discuss mailing list > ksh93-integration-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/ksh93-integration-discuss > -- , _ _ , { \/`o;====- Olga Kryzhanovska -====;o`\/ } .----'-/`-/ olga.kryzhanov...@gmail.com \-`\-'----. `'-..-| / Solaris/BSD//C/C++ programmer \ |-..-'` /\/\ /\/\ `--` `--` _______________________________________________ ksh93-integration-discuss mailing list ksh93-integration-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/ksh93-integration-discuss
