On Mon, 30 Dec 2019 09:33:48 +0900
"Michele Calgaro via trinity-devel"
<trinity-devel(a)lists.pearsoncomputing.net> wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Hi E.
[1010/2200] cd
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999_build/doc/tdespell
&&
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999_build/kdoctools/meinproc
--srcdir=/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999/kdoctools --check
--cache index.cache.bz2
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999/doc/tdespell/index.docbook
FAILED:
doc/tdespell/index.cache.bz2 cd
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999_build/doc/tdespell
&&
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999_build/kdoctools/meinproc
--srcdir=/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999/kdoctools --check
--cache index.cache.bz2
/var/tmp/portage/trinity-base/tdelibs-9999/work/tdelibs-9999/doc/tdespell/index.docbook
This could be related to the creation of index docbook. I had a simial issue many years
ago with meinproc. If I remember
correctly, it was something related to the calling of tmoc before meinproc was called.
Don't remember the exact details
though.
Probably not anything I did, then. Fixing it presumably requires patching part of the
build system—a bit beyond me at this time.
A comparison
of multiple versions of CMakeLists.txt shows that the WITH_UDISKS and WITH_UDISKS2 options
are being
removed. Are they supposed to be subsumed into WITH_TDEHWLIB, or is udisks2 going to be
an absolute requirement
for building TDE in the future? (Minimalist configurations are quite common in Gentoo,
so I need to get this
right.)
In R14.1.x, tdehw lib will try to use any of the available methods in a system to
mount/umount disks.
The order of testings is udisks2, udisks, udevil, pmount. Calls are done through dbus
(udisks2, udisks) or thorugh popen
(udevil, pmount).
The options WITH_UDISKS and WITH_UDISKS2 have been removed because now the code is always
compiled and if udisks/udisks2
is not available at runtime, the calls will fail and the next method will be tried out.
Since the call are done through
DBUS, there is no hard dependency at compile time.
So WITH_TDEHWLIB requires dbus and one of udisks2, udisks-not-2, udevil, or pmount.
(Note that my attempted tdelibs git build blew up earlier and somewhere in tdehwlib before
I
set WITH_TDEHWLIB to "OFF", so I'm not sure there isn't an inadvertant
build-time
dependency on udisks and pals in the code somewhere. The chroot I'm building in has
dbus
installed, but none of the others.)
Do any of the other options (such as the encryption-related ones below) depend on
tdehwlib being enabled?
There are
three new options: WITH_PCSC, WITH_PKCS, and WITH_CRYPTSETUP, one of which has a
not-very-useful
description. My understanding is that:
These options were added by Tim 5 years ago or so and actually some are a bit of an issue
for R14.1.x.
WITH_PCSC offers support for smartcards and
requires the pcsc-lite library
Correct, AFAICT.
Okay, so I'll need to add a USE flag "pcsc-lite" to the ebuild to support
switching build support
for this feature on and off. And dig up the documentation on local USE flags in
overlays.
WITH_PKCS
offers support for PKCS#11 crypto and requires (I think) the library packaged by Gentoo
as
dev-libs/pkcs11-helper
Yes, require pkcs11-helper and AFAICR it does not support ssl 1.1 (haven't checked
for long time) so in R14.1.x we have
disable this option at build time for the time being
I'll also hard-code this one off for the time being, then.
WITH_CRYPTSETUP offers support for filesystem encryption(?) using sys-fs/cryptsetup(?)
cryptsetup is related to support for LUKS (and plain) encrypted disks. If this is
disabled, you should lose the
capability of being able to modify properties of LUKS disks (in particular keys) from the
GUI. Anyhow it should not
affect the capability to mount/umount LUKS disks at runtime as long as you have
cryptsetup package installed, since
those operations are once again managed by DBUS calls.
Please note that the code in R14.1.x master branch related to LUKS disk support does not
work properly. I have a PR open
for it where I have already made lot of work. I just need to find a spell of time where I
can work on it with some
consistency before I can merge it to master.
Add USE flag "cryptsetup" for this one, then, but mask it for the time being to
prevent it
being turned on before the feature is fixed. (Talking partly to myself here . . .)
E. Liddell