Debian Patches
Status for lrzsz/0.12.21rc-4
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| ansi2knr.diff | Remove ansi2knr rules | Bastian Germann <bage@debian.org> | not-needed | |||
| gettext_external.diff | Call AM_GNU_GETTEXT with external | Bastian Germann <bage@debian.org> | not-needed | |||
| ignore-intl.diff | Ignore building intl | Bastian Germann <bage@debian.org> | not-needed | |||
| 313992_de_po.diff | Update the German message catalogue (po/de.po) | Tobias Toedter <t.toedter@gmx.net> | not-needed | debian | ||
| build.diff | Drop AC_PREREQ([2.13]) from the bundled automake macros | not-needed | ||||
| cat-id-tbl.diff | Refresh the generated po/cat-id-tbl.c message table | not-needed | ||||
| include.diff | Add missing system includes (stdlib.h) | not-needed | ||||
| mantypos.diff | Fix a groff macro typo in lsz.1 (missing space after .B) | not-needed | ||||
| strstr.diff | Drop the conflicting extern strstr() declaration | not-needed | ||||
| 897010_zsdata.diff | Fix zsdata() zero-length data subpacket handling | not-needed | debian | |||
| 1066514_FTBFS_implicit_declaration.diff | Fixes FTBFS (missing includes) | Martin A. Godisch <godisch@debian.org> | not-needed | debian | ||
| 206648_DSZLOG_support.diff | DOS-style DSZLOG support It is useful when using lrzsz with BBS software packages. | Ryan Underwood <nemesis@dbz.icequake.net> | not-needed | debian | ||
| 1037046_typo_manpage.diff | Fixed typo in manpage | Uwe Kleine-König <ukleinek@debian.org> | not-needed | debian | ||
| warnings.diff | fixes some compiler warnings | Martin A. Godisch <godisch@debian.org> | not-needed | |||
| configure.in.diff | Remove de-ANSI-ficcation | Bastian Germann <bage@debian.org> | not-needed | |||
| gettext_require_version.diff | Add AM_GNU_GETTEXT_REQUIRE_VERSION to fix autopoint FTBFS configure.in calls AM_GNU_GETTEXT (turned into AM_GNU_GETTEXT([external]) by gettext_external.diff) but never declares a gettext infrastructure version. Modern autopoint (gettext >= 0.18; sid currently ships 0.26) then falls back to the pre-versioning 0.10.x scheme, requests the long-removed gettext-0.10.32 layout and aborts autoreconf with "requires the infrastructure from gettext-0.10.32 but this version is older". Declaring a modern minimum via AM_GNU_GETTEXT_REQUIRE_VERSION([0.19.8]) gives autopoint a valid version to work from; the newer gettext installed on the build host is used at build time. |
Martin A. Godisch <godisch@debian.org> | not-needed | 2026-06-04 | ||
| po_makevars_linguas.diff | Add po/Makevars and po/LINGUAS for the modern gettext infrastructure Declaring AM_GNU_GETTEXT_REQUIRE_VERSION makes autopoint replace the 1998-era po/Makefile.in.in with the one from the host's gettext (>= 0.19). That modern Makefile.in.in has a build rule depending on po/Makevars, which this package never shipped, so make aborts with "No rule to make target 'Makevars'". . po/Makevars carries the canonical gettext variables (message domain, xgettext keywords, copyright holder, ...). po/LINGUAS lists the translations to build; without it the modern infrastructure reads an empty language set and silently drops the existing German catalogue (de.po), so it is added with "de". |
Martin A. Godisch <godisch@debian.org> | not-needed | 2026-06-04 | ||
| compiler_warnings.diff | Fix the remaining compiler warnings in the C sources Targets the warnings that survived warnings.diff, all in src/: . * lsyslog.c (-Wformat=): the dszlog() fprintf() uses %ld conversions but passed plain int arguments (cps*8, cps and the -1 serial-number literal). Passing an int where the format expects a long is undefined behaviour in a variadic call and prints garbage on ABIs where int and long differ in size. Cast those three arguments to long. * lsyslog.c, lrz.c, zm.c, canit.c, lsz.c (-Wunused-result): the return values of write(), system() and vasprintf() were discarded, which the fortified glibc headers flag. Consume them explicitly: the best-effort write()/system() calls are wrapped in "if (... ) { }" guards, and on a vasprintf() failure the target pointer is set to NULL so that the later syslog() and free() operate on a well-defined value. . No functional change beyond the vasprintf() NULL safeguard. |
Martin A. Godisch <godisch@debian.org> | not-needed | 2026-06-04 | ||
| 1105467_shuffle_install_order.diff | Fix FTBFS with make --shuffle=reverse automake wires install-exec-local into install-exec-am next to install-binPROGRAMS but does not order the two. install-exec-local hard-links the already-installed binaries (sz, rz), so it only works because make happens to honour the prerequisite list order. With make --shuffle=reverse it runs before install-binPROGRAMS, the link targets do not exist yet and ln aborts with "failed to access '.../usr/bin/sz'". Make the ordering explicit. |
Martin A. Godisch <godisch@debian.org> | not-needed | debian | ||
| 57884_syslog_locale_timestamp.diff | Keep syslog timestamps locale-independent lrz and lsz call setlocale(LC_ALL, "") so that gettext picks up the user's language. That also sets LC_TIME, which the C library's syslog() honoured when formatting the per-message timestamp, so sz/rz lines in the system log carried a translated month name (e.g. the Bulgarian month abbreviation in #57884) while every other line stayed in English. Only message translation needs the environment locale, so reset LC_TIME back to "C" right after; this leaves gettext and LC_CTYPE untouched and has no effect on the programs' own output (their sole strftime use is the numeric "%H:%M:%S"). |
Martin A. Godisch <godisch@debian.org> | not-needed | debian | ||
| 156249_xmodem_block_counter.diff | Fix the XMODEM "Blocks received" counter past block 255 wcrx() reported wcgetsec()'s return value, i.e. the raw one-byte XMODEM block number, which wraps 255 -> 0 -> 1. From block 256 on the counter therefore restarted at 0; because report() redraws the line with a bare carriage return, the shorter new number left the trailing digits of the previous one on screen, so "256" showed up as "055", "257" as "155" and so on. Report the running, non-wrapping sector total (sectnum) instead, and only once a block has actually been accepted. The counter now climbs monotonically (...255, 256, 257, ...), which also removes the stale-digit artefact since its width never shrinks. |
Martin A. Godisch <godisch@debian.org> | not-needed | debian | ||
| fix_siginterrupt_deprecation.diff | Replace deprecated siginterrupt() with sigaction() siginterrupt(SIGALRM, 1) in lrz.c/lsz.c flagged the C library so that the SIGALRM handlers later installed with signal() in zreadline.c and tcp.c would not set SA_RESTART, i.e. the timeout alarm interrupts the blocking read()/accept()/connect() instead of restarting it. siginterrupt() is now deprecated (-Wdeprecated-declarations). Install those three handlers directly with sigaction() and sa_flags = 0, which expresses the same "do not restart" semantics locally, and drop the now-unneeded siginterrupt() calls. No behavioural change; sigaction() is POSIX and always available on Debian. |
Martin A. Godisch <godisch@debian.org> | not-needed | |||
| 618357_manpage_remove_test_mode.diff | Drop documentation of the non-existent sz -T/-TT test modes The lsz.1 manual page still described two test/diagnostic modes inherited from Chuck Forsberg's original sz: a "sixth form" sz -TT that supposedly dumps all 256 byte codes to the terminal, and a "TESTING FEATURE" in which sz -T file exercises the Attn error-recovery sequence. Neither exists in lrzsz: -T is bound solely to --turbo (turbo_escape). A user following the manual therefore ran "sz -TT", which is just --turbo twice, and got a normal (failing) transfer to the terminal instead of a diagnosis (Debian #618357). Remove the stale -TT synopsis form, the "sixth form" paragraph, the "TESTING FEATURE" section and the orphaned BUGS sentence so the page matches the actual behaviour. |
Martin A. Godisch <godisch@debian.org> | not-needed | debian | ||
| 206499_ymodemg_doc.diff | Clarify in lsz.1 that YMODEM-g must be requested by the receiver The sb manual page advertises YMODEM-g support, which led to the report that the feature was "nowhere to be found" (Debian #206499): YMODEM-g is in fact implemented on the send side (sb streams without per-sector ACKs once the receiver sends 'G' instead of the usual start character), but the receiver shipped with lrzsz (rb/rz) cannot request it, so a user driving both ends with this package never sees it. Add a note to the sb description spelling out that sb never enables YMODEM-g on its own and that lrzsz's own rb does not request it, so a separate receiving program is required. Documentation only. |
Martin A. Godisch <godisch@debian.org> | not-needed | debian |
All known versions for source package 'lrzsz'
- 0.12.21rc-4 (forky, sid)
- 0.12.21rc-0.2 (trixie)
- 0.12.21-10 (bookworm)
