| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| fuzz-link | commit 4d3160a8a6fb571bfaaf540f57a32c99b44b6fa7 fuzz: Use main from libfuzzer diff --git a/Makefile.am b/Makefile.am index 3b002e4d..6d7e5764 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-15 | ||
| th_dict | commit 315fec6690fc9f86c5e1a40c41d29025d4b7db55 Fix generating th_dict diff --git a/Makefile.am b/Makefile.am index 6d7e5764..cbe04b59 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-15 | ||
| klatt-garbage | commit 7e304a41295fa6b4458f759e8ac70f86b61b26e5 klatt: Fix out-of-bound access klatt should watch for wdata->mix_wavefile_max like others do already. Probably we should actually be scaling the wave according to speed.wav_factor rather than backwarding by 3/4, but for now let's at least fix the garbage introduced in the klatt case (which was also getting noticed on big-endian archs where the garbage is different). diff --git a/src/libespeak-ng/klatt.c b/src/libespeak-ng/klatt.c index dfadcfd9..db389199 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-15 | ||
| libsonic-bigendian | commit c794ef2c9a868e3d901b637f35ef90bd16566ff1 libsonic: Fix support on big-endian archs libsonic reads/writes in host order while espeak-ng works with little-endian order. diff --git a/src/libespeak-ng/wavegen.c b/src/libespeak-ng/wavegen.c index eb00106f..07e44f57 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-16 | ||
| bufsize | Apparently 60ms is too much for some cards. | no | ||||
| compile-reproducibility | commit 98321c9d281843bf62fd59c9872d2ee2e486f98e compiledict: make rgroup_sorter not depend on allocator To get reproducibility, we should not depend on the memory allocator-provided pointer values. We can instead just use the group creation order. diff --git a/src/libespeak-ng/compiledict.c b/src/libespeak-ng/compiledict.c index 9304f4e1..7fb1fb79 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-15 | ||
| espeak-stdin | commit 33b6daf420710eca2f6b2dcdc8eb3f6d72e1d313 Fix dropping last byte of stdin input This got broken with eaa0c9aa08350ce8a1343aed4c0a87c1cc329b0e ("Fix truncated fgetc return value in main(espeak-ng.c). [Coverity]") which rightfully stopped adding a bogus EOF at the end of the input, which this line was probably dropping just by luck. Fixes brailcom/speechd#978 diff --git a/src/espeak-ng.c b/src/espeak-ng.c index ee63da70..2ea3791a 100644 |
Samuel Thibault <samuel.thibault@ens-lyon.org> | no | 2024-12-21 | ||
| fix-cancel | https://github.com/espeak-ng/espeak-ng/issues/172 Upstream pcaudiolib made audio_object_flush actually do the close/reopen, but this is not released yet (will be in 1.1). This however poses threading issues with ALSA: audio_object_flush is currently called in the application thread while it must be done in the say_thread for proper safety. espeak-ng 1.49.2 added a cancel_audio() function which calls audio_object_flush and is called from close_stream(). That is however not enough for the stream to get interrupted on cancel (and we don't want to use upstream's current fix mentioned above since it poses threading issues). |
no | ||||
| privacy | no | |||||
| x | Avoid complaining about our own files being executable :) | no | ||||
| CI | CI on salsa seems to be detecting an exec bit ?? | no | ||||
| nocard | no | |||||
| clang-target | https://github.com/espeak-ng/espeak-ng/pull/2063 | no |