This was disabled in 9ff1828308 due to
buggy pthread support, but this this option does not induce sqlite
to use multiple threads. Instead, it builds sqlite with the
necessary mutexes to make the library thread safe.
https://sqlite.org/threadsafe.html
Most of the automatic conversion was done using the following Python script:
import os
import re
pkgs = sorted(mkfile[:-len('.mk')] for mkfile in os.listdir('src') if mkfile.endswith('.mk'))
with open('index.html', 'rb') as f:
index = f.read()
pkgs_index = re.findall('<td class="package">([^<]*)</td>', index)
assert pkgs_index == pkgs
versions = dict(re.findall('<td id="([^"]*)-version">([^<]*)</td>', index))
assert sorted(versions.keys()) == pkgs
for pkg in pkgs:
version = versions[pkg]
with open('src/' + pkg + '.mk', 'rb') as f:
mk = f.read()
checksumpos = mk.index('\n$(PKG)_CHECKSUM ')
versionline = '\n$(PKG)_VERSION := %(version)s' % {'version': version}
newmk = mk[:checksumpos] + versionline + mk[checksumpos:]
with open('src/' + pkg + '.mk', 'wb') as f:
f.write(newmk)
Unsetting CONFIG_SITE was the right way to solve the problem that
--libdir was working around.
This reverts commit 0b3240f9d8.
This reverts commit 3653eb6ec1.
This reverts commit 593ab5a706.
For now this means a bit of duplication, but if the remaining
references to SUBDIR can be dealt with, we should be able
to get rid of the SUBDIR defintion altogether.