Thread: Procedure of modification of autoconf scripts
Hi all,
Can you please explain the process of adding new items into autoconf scripts? Specifically into configure.ac. For example, if I want to add a new --with-foo argument, let's say a new 3rd party library. What should I do after proper configure.ac modification? Should I also re-generate configure script with local autoreconf? My doubts are that changes to configure script can be rather huge and likely conflicting with other patches that possibly do the same. Thanks!
Ian Zagorskikh <izagorskikh@cloudlinux.com> writes:
> Can you please explain the process of adding new items into autoconf
> scripts? Specifically into configure.ac. For example, if I want to add a
> new --with-foo argument, let's say a new 3rd party library. What should I
> do after proper configure.ac modification? Should I also re-generate
> configure script with local autoreconf? My doubts are that changes to
> configure script can be rather huge and likely conflicting with other
> patches that possibly do the same. Thanks!
If you see massive changes in the configure script after a localized
change in configure.ac, it probably means that you're not using the
right autoconf version.
Our project convention is to use exactly the GNU release of whichever
version of autoconf we're on (currently 2.69). A lot of vendors ship
modified-to-some-extent autoconf versions, which can result in these
sorts of unwanted changes if you just use whatever is on your
operating system. Grab the official release off a GNU mirror and
install it somewhere handy, and use that.
As a test case, try running autoconf and autoheader *without*
having changed the input files. If the outputs don't match
what's in git, then you've got something to fix.
regards, tom lane
Tom,
Thank you now it's clear!
On Thu, Apr 22, 2021 at 5:58 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
If you see massive changes in the configure script after a localized
change in configure.ac, it probably means that you're not using the
right autoconf version.
Our project convention is to use exactly the GNU release of whichever
version of autoconf we're on (currently 2.69). A lot of vendors ship
modified-to-some-extent autoconf versions, which can result in these
sorts of unwanted changes if you just use whatever is on your
operating system. Grab the official release off a GNU mirror and
install it somewhere handy, and use that.
As a test case, try running autoconf and autoheader *without*
having changed the input files. If the outputs don't match
what's in git, then you've got something to fix.
regards, tom lane
--