You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@apr.apache.org by "William A. Rowe, Jr." <wr...@rowe-clan.net> on 2001/02/26 01:24:59 UTC
The list...
Jeff... thanks for banging your head against 2.0 for us all.
I took a look at the 'private' old HAVE_* subset, listed below;
HAVE_DL_H
HAVE_DLFCN_H
HAVE_GRP_H
HAVE_HSTRERROR
HAVE_ICONV_H
HAVE_LANGINFO_H
HAVE_PWD_H
HAVE_MALLOC_H
HAVE_STDDEF_H
HAVE_SYS_FCNTL_H
HAVE_SYS_FILE_H
HAVE_SYS_MMAN_H
HAVE_SYS_RESOURCE_H
HAVE_SYS_SELECT_H
HAVE_SYS_SEM_H
HAVE_SYS_STAT_H
HAVE_TIME_H
HAVE_ICONV
HAVE_LIBDL
HAVE_POLL
HAVE_PTHREAD_KEY_DELETE
HAVE_WRITEV
I ment to back out all of these, from your last commit, I take it I failed
to do so. I'm sorry about that.
The question is which of these are worth exporting publicly. I don't have
a fast or simple answer. So I'm just tossing the list at the list.
Bill
Re: The list...
Posted by rb...@covalent.net.
On Sun, 25 Feb 2001, Wan-Teh Chang wrote:
> cliffwoolley@yahoo.com wrote:
> >
> > Fair enough. So let me tone down what I said to simply, "If an APR client
> > might reasonbly need the same test, we should export our test result."
> > +1 for using Apache and Subversion as good places to look for an initial
> > definition of what's "reasonable."
>
> If an APR client needs the same test, you might want
> to look into why APR fails to make the test unnecessary
> for its clients. For example, the apr_dso_* functions
> should make the tests for HAVE_DL_H and HAVE_DLFCN_H
> unnecessary for APR clients.
>
> For those HAVE_* macros that really need to be exported,
> you might want to rename them APR_HAVE_* to stay inside
> the APR_ namespace.
We have already done the APR_HAVE_* rename, because otherwise we collide
with Apache's autoconf checking. Most of the code in Apache that needs
autoconf macros is code that doesn't use APR for one reason or another.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: The list...
Posted by cl...@yahoo.com.
On Sun, 25 Feb 2001, Wan-Teh Chang wrote:
> If an APR client needs the same test, you might want
> to look into why APR fails to make the test unnecessary
> for its clients. For example, the apr_dso_* functions
> should make the tests for HAVE_DL_H and HAVE_DLFCN_H
> unnecessary for APR clients.
That's true to an extent... apr_want.h addresses many such issues. But at
the same time, some clients are just going to want to do things in ways we
haven't 100% predicted. If we've already done the test in APR, we might
as well export the results to save the client a little effort. In other
words, we give them as many tools as possible (eg apr_want.h) for them not
to need the macros at all, but if they do want them and we already know
their value, we might as well provide them.
> For those HAVE_* macros that really need to be exported,
> you might want to rename them APR_HAVE_* to stay inside
> the APR_ namespace.
Oh, clearly... I assumed that that went without saying. That's definitely
the plan. And moreover, if APR_HAVE_* exists, it is always defined (to 0
or 1), as opposed to plain-old autoconf HAVE_* macros which are either
defined or not.
--Cliff
Re: The list...
Posted by Wan-Teh Chang <wt...@netscape.com>.
cliffwoolley@yahoo.com wrote:
>
> Fair enough. So let me tone down what I said to simply, "If an APR client
> might reasonbly need the same test, we should export our test result."
> +1 for using Apache and Subversion as good places to look for an initial
> definition of what's "reasonable."
If an APR client needs the same test, you might want
to look into why APR fails to make the test unnecessary
for its clients. For example, the apr_dso_* functions
should make the tests for HAVE_DL_H and HAVE_DLFCN_H
unnecessary for APR clients.
For those HAVE_* macros that really need to be exported,
you might want to rename them APR_HAVE_* to stay inside
the APR_ namespace.
Re: The list...
Posted by cl...@yahoo.com.
On Sun, 25 Feb 2001 rbb@covalent.net wrote:
> > At least all of the HAVE_foo_H ones, I'd think. If there's a reasonable
> > chance an APR client will need to do the same autoconf test, we might as
> > well save them the trouble. Header tests pretty easily fall into that
> > category.
>
> I disagree, but only a little. :-) Some of those header checks are for
> information that APR completely replaces. We check for the headers, so
> that we can use the headers, but we don't expect APR applications to use
> them at all. I do believe that any header Apache or Subversion is
> currently using is a good place to start however.
Fair enough. So let me tone down what I said to simply, "If an APR client
might reasonbly need the same test, we should export our test result."
+1 for using Apache and Subversion as good places to look for an initial
definition of what's "reasonable."
--Cliff
Re: The list...
Posted by rb...@covalent.net.
On Sun, 25 Feb 2001 cliffwoolley@yahoo.com wrote:
> On Sun, 25 Feb 2001, William A. Rowe, Jr. wrote:
>
> > The question is which of these are worth exporting publicly. I don't have
> > a fast or simple answer. So I'm just tossing the list at the list.
>
> At least all of the HAVE_foo_H ones, I'd think. If there's a reasonable
> chance an APR client will need to do the same autoconf test, we might as
> well save them the trouble. Header tests pretty easily fall into that
> category.
I disagree, but only a little. :-) Some of those header checks are for
information that APR completely replaces. We check for the headers, so
that we can use the headers, but we don't expect APR applications to use
them at all. I do believe that any header Apache or Subversion is
currently using is a good place to start however.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: The list...
Posted by cl...@yahoo.com.
On Sun, 25 Feb 2001, William A. Rowe, Jr. wrote:
> The question is which of these are worth exporting publicly. I don't have
> a fast or simple answer. So I'm just tossing the list at the list.
At least all of the HAVE_foo_H ones, I'd think. If there's a reasonable
chance an APR client will need to do the same autoconf test, we might as
well save them the trouble. Header tests pretty easily fall into that
category.
--Cliff