Debian Patches

Status for speech-tools/1:2.5.0-13

Patch Description Author Forwarded Bugs Origin Last update
debian_build.diff This patch provides most of the patches for speech-tools for Debian GNU/Linux,
to provide Debian specific building of speech-tools.
no
config_config.in.diff This patch is part of a series of patches for speech-tools for
Debian GNU/Linux, to provide Debian specific building of speech-tools.

This patch was not forwarded upstream, since it is Debian specific. However,
a related patch was accepted (https://github.com/festvox/speech_tools/pull/30). On
the next upstream release, consider removing this patch and changing the debian/rules
so they use `./configure SHARED=2 DEBUG=1` instead.
no
datadir_fhs.diff On festival and speech-tools 1.4, arch-independent files were stored on datadir,
while arch-dependent files went to libdir. After that version, libdir and datadir
were merged upstream into libdir, breaking FHS conventions.

This patch reverts that merge, so libdir has the architecture-dependent files and datadir
has the architecture-independent files.

All festival dependencies follow upstream festival (using libdir for everything) and in
Debian we patch them to use datadir when required.

As far as I (Sergio Oller, 2020) know, the only thing that actually goes to libdir and that
it is architecture-dependent is /usr/lib/festival/audsp, from the festival package. The
rest of references to libdir are changed to datadir. This involves heavy patching.

Maybe it would be worth considering removing this patch, and making
libdir point to /usr/share instead of /usr/lib.

Any use of architecture-dependent files in libdir would need to be patched (such as audsp).
But that would require extremely minor patching in comparison to the current state.
no
use_standard_headers.diff This patch is for speech-tools for Debian GNU/Linux,
so as to use the standard library rather than its older form.
no
include_EST_TIterator.h.diff This patch is for speech-tools for Debian GNU/Linux,
and forces the use of the actual template rather than that which
is inherits from when defining Iter typedef.
no
alaw.diff Allow alaw output (asterisk interoperability) no
pulseaudio_and_alsa.diff yes
Makefile.diff This patch is part of a series of patches for speech-tools for
Debian GNU/Linux, to provide Debian specific building of speech-tools
no
copyright_4clausefix.diff This patch removes the advertising clause from the BSD project's
files, as the BSD project has deprecated the 4-clause license with
that clause anyway.
no
lintianspell.diff Correct a spelling error flagged by lintian Peter Drysdale <drysdalepete@gmail.com> yes
templatefixes.diff This patch adds the not-equal operator for EST_TList.
This patch is not intended to alter the code in a substantive way.
No copyright subsists in this patch as it is too trivial.
Peter Drysdale <drysdalepete@gmail.com> no
dontbuildesd.diff Don't build EsounD support since it was deprecated years ago
by Gnome. This is a Debian build system specific patch.

As far as I (Sergio Oller, 2020) this patch is not needed since the
ESD_AUDIO module is already disabled in config/config.in
"# INCLUDE_MODULES += ESD_AUDIO".
no
tokenstream_istream.diff This patch provides a seeking and reading interface to istreams in the
EST_TokenStream class.
This patch is required by the htsperformance.diff patch in Festival
in order to reduce the i/o disk usage when using HTS voices.
This patch is not intended to alter the code in a substantive way.
No copyright subsists in this patch as it is too trivial.
Sergio Oller <sergioller@gmail.com> no
wave_save_parts.diff This patch provides methods to EST_Wave for saving wave files
by parts, either writing only the file header or the actual data.
Additionally it provides a function to write a wave file header without
having an EST_Wave instance.

This patch can be used (for instance by festival). On Festival-2.1 the way
text2wave works is as follows: Given a long text, it is split in utterances
which are saved in temporal files. At the end, all temporal files are loaded
into memory and appended in memory one by one, and finally dumped to the
output file.
This approach required double the disk i/o (temporal files and final file) and
required to have the whole synthesized text loaded in memory.

By using this patch, text2wave can be improved so as utterances are written
directly to the final file reducing both disk i/o and memory usage at once.

Backwards compatibility is preserved.

No copyright is claimed on this patch, in order to make it easier to
merge with upstream, although attribution is appreciated :-)
Sergio Oller <sergioller@gmail.com> no
wave_save_parts_alaw.diff This patch complements wave_save_parts and the alaw support.
It allows one to save raw data in alaw format using the same interface
than the one provided in wave_save_parts.

This patch depends on wave_save_parts.diff and the alaw patches to
provide asterisk support.

No copyright subsists in this patch as it is too trivial.
Sergio Oller <sergioller@gmail.com> no
check_fread.diff no
memory_bugs.diff Modify data loading memory management to avoid crashes.This patch fixes a crash in audsp in Festival and other
memory leaks.
Samuel Thibault. no upstream
ch_wave-LONGEST.diff This patch allows providing "-pc LONGEST" argument in lowercase.
Bug reported by chaered: https://bugs.launchpad.net/ubuntu/+source/speech-tools/+bug/723854
Sergio Oller <sergioller@gmail.com> no
hardening.diff This patch allows one to build speech tools honoring user provided
CFLAGS, CXXFLAGS and LDFLAGS. This is required to have a hardened build
without the obsolete hardening-wrapper package.

The patch also fixes config/rules/test_make_rules.mak so the test binaries
also respect LDFLAGS when being linked. This change fixes the Debian QA buildd
check and has already been submitted upstream at
https://github.com/festvox/speech_tools/pull/5/files

Last update: 2018-07-19
Sergio Oller <sergioller@gmail.com> no
timestamp no
more-lintian-spelling.patch Lintian reports typos Paul Gevers <elbrus@debian.org> yes
link_estools_openmp.diff This patch adds OpenMP flags to the command used to link the shared library.
OpenMP is used by the wagon tool since speech-tools-2.5. It is used in the libestools
shared library. dpkg-shlibdeps reports missing OpenMP symbols if the shared library is not
linked with OpenMP options.

Last update: 2018-06-30
Sergio Oller <sergioller@gmail.com> no
test_failure make make test fail when the testsuite fails Samuel Thibault <sthibault@debian.org> no
no-common no
reproducible.diff Avoid recording build flags in a file, we can get them from the environment.

https://github.com/festvox/speech_tools/pull/46
no

All known versions for source package 'speech-tools'

Links