minicom: update to use latest git repository

Minicom is now available on a git repository at:
https://salsa.debian.org/minicom-team/minicom.git
Lot of fixes has been added since version 2.7-1 and RS485 support has been added too under serial port configuration.

Git shortlog follows starting from version 2.7 since commit for 2.7-1 debian version is not present.
The only difference between version 2.7-1(debian) and 2.7 seems to be a coding style change in src/config.c file that breaks every line into a string like:
(
"aaa"
"bbb"
"ccc"
)
instead of:
(
"aaa\
bbb\
ccc"
)
**************************************************************************
Adam Lackorzynski (85):
Added tag v2.7 for changeset ca0e6c0e1c7d
Set version to 2.7.90
Rename configure.in to configure.ac
Add delta timestamp mode
Enable SIGALRM signal
Implement HPA ESC sequence
Debugging aid (disabled)
man: add "U" key shortcut description
Fix filename of non-global config settings.
Update indonesian translation
Update french translations
Use dial_tty in statusline
Update by french translation team
Add more TODO/ideas
musl-libc compile fix
Change MAXNAMLEN to POSIX's NAME_MAX
Convert ChangeLog to UTF8
Add alternative window support
Remove SMOOTH code blocks
da.po: Update from translation team
Add ARRAY_SIZE macro
Add default for missing function keys
TODO: move location of config files to ~/.config
Add F11+F12 for use in macros
nb.po: Update from translation team
script: Proper shell exit status handling.
Add bug description
Improvement to socket handling code
autogen.sh: Set to 1.15
loadconv: Add missing fclose()
Add DTR toggle via TIOCMBIC+TIOCMBIS
remove conditional DTR toggle enable
Add entry
Discontinue Changelog
Fix printf warning on 64bit
Updated swedish translation by Sebastian Rasmussen
Fix color setting vt_pinit
Color settings: Simplify same term color message
Add "bug fixes" entry
Change mbslen to mbswidth
Add check_io_input() function
Fix return value when socket got closed
Add serbian translation
TODO: Transmit delay has no influence on macros
TODO: Need to move .macros location too
Use vt_send for macros too
Improve status line redrawing
Simplify pointer logic in output function
Fix CVE-2017-7467, a remote exploitalbe out of bound access
TODO: Add to make literal capture available
TODO: Add note on not running as root
Remove absolute paths to called programs
Make upload/download window wider
windiv: Add sys/types.h
Add sys/wait.h include
Add a missing 'free' for a calloc.
Fix type of vt_tabs for proper left shift
Free dialents on exit
dial: Remove tagged variable and code
Use ARRAY_SIZE macro
Zero initialize malloc'ed buffer
Fix behind-buffer access in input window
Fix buf size calculation by using ARRAY_SIZE
Fix value for one mbswidth test
Fix writing over end of array
Redraw status line on screen size changes
Support a second value for the mbswidth check
Fix build with SVR4_LOCKS
Fix a -Wformat-truncation warning from gcc7
Fix -Wformat-truncation warning by gcc7
Annotate switch-case fall-throughs
Fix parsing of MINICOM env-var
Introduce -O option, and add timestamps
timestamping: Only print if prev time exists
Add simplified chinese translation
man: Add current maintainer
Initial support for TCP connections
Abstract check_io functionality
Change return type of check_io_input to bool
Update spanish translation
Flush output before delaying output
ascii-xfr: Put flushing into a function
ascii-xfr: Apply 'static'
configure.ac: Add -Wno-format-truncation
Common FALLTHRU style

Giulio Benetti (4):
Add RS485 parameters.
Add RS485 menu entries to "Serial port Setup" Handle RS485 low-level
sysdep1: avoid asserting rts when rs485 enabled

Martin A. Godisch (2):
fixed text width in history buffer message box properly escaped dashes in manual page

Matthias Dieter Walln?fer (1):
Properly recognise multi-byte characters by introducing

Ond?ej Lyson?k (6):
Add a missing va_end() call
Make sure strings copied by strncpy are null-terminated Fix file descriptor leaks
Fix a directory handle leak
Fix a read past end of buffer
Fix a warning about an unused variable

Sebastian Rasmussen (3):
src/man: Fix typos in documentation and comments Fixes #315391
src: Use gettext in ascii-xfr to make it translatable Fixes #315392
src: Make various messages/strings translatable Fixes #315392

**************************************************************************

Here is shortlog for Buildroot:

Giulio Benetti (4):
minicom: update package to use git repository.
minicom: add pre-configure hook to call autogen.sh
minicom: delete patches
minicom: delete hash file.

 .../minicom/0001-musl-libc-compile-fix.patch  | 28 ----------
 ...-change-maxnamlen-to-posixs-name-max.patch | 51 -------------------
 package/minicom/minicom.hash                  |  2 -
 package/minicom/minicom.mk                    | 14 ++++-
 4 files changed, 12 insertions(+), 83 deletions(-)
 delete mode 100644 package/minicom/0001-musl-libc-compile-fix.patch
 delete mode 100644 package/minicom/0002-change-maxnamlen-to-posixs-name-max.patch
 delete mode 100644 package/minicom/minicom.hash

2.17.1

Minicom is now hosted on a git repository:
https://salsa.debian.org/minicom-team/minicom.git

Change package to use git method pointing to last commit.

---
 package/minicom/minicom.mk | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

index cb06482379..1a3426899f 100644
@@ -4,8 +4,9 @@
 #
 ################################################################################
 
-MINICOM_VERSION = 2.7.1
-MINICOM_SITE = https://alioth.debian.org/frs/download.php/file/4215
+MINICOM_VERSION = 0e79d0bb731a16a3126d926740f923bcb6ecd738
+MINICOM_SITE = https://salsa.debian.org/minicom-team/minicom.git
+MINICOM_SITE_METHOD = git
 MINICOM_LICENSE = GPL-2.0+
 MINICOM_LICENSE_FILES = COPYING
 

2.17.1

Using latest minicom git commit need to call autogen.sh instead of relying on Buildroot MINICOM_AUTORECONF = YES

Add MINICOM_RUN_AUTOGEN to MINICOM_PRE_CONFIGURE_HOOKS and dependencies needed by autogen.sh like: host-automake, host-autoconf, host-libtool.

---
 package/minicom/minicom.mk | 9 +++++++++
 1 file changed, 9 insertions(+)

index 1a3426899f..91eb74c9ad 100644
@@ -18,4 +18,13 @@ MINICOM_CONF_ENV = PKG_CONFIG=/bin/false
 
 MINICOM_DEPENDENCIES = ncurses $(if $(BR2_ENABLE_LOCALE),,libiconv)
 
+# configure is missing and minicom is not compatible with our autoreconf
+# mechanism so we have to do it manually instead of using MINICOM_AUTORECONF = YES
+define MINICOM_RUN_AUTOGEN
+	cd $(@D) && PATH=$(BR_PATH) ./autogen.sh
+endef
+MINICOM_PRE_CONFIGURE_HOOKS += MINICOM_RUN_AUTOGEN
+
+MINICOM_DEPENDENCIES += host-automake host-autoconf host-libtool
+
 $(eval $(autotools-package))

2.17.1

Hello,

Looking at the autogen.sh script, I don't see anything fancy. What is happening when you use MINICOM_AUTORECONF = YES ?

Best regards,

Thomas

Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com

Hello,

Il 01/08/2018 16:48, Thomas Petazzoni ha scritto:

It complains about m4 directory missing.
Maybe I could create a workaround for that?

Giulio

Hello,

Example:

# Autoreconf requires an existing m4 directory
define BLUEZ_ALSA_MKDIR_M4

        mkdir -p $(@D)/m4

endef
BLUEZ_ALSA_POST_PATCH_HOOKS += BLUEZ_ALSA_MKDIR_M4

Best regards,

Thomas

Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com

Hi,

Il 01/08/2018 17:10, Thomas Petazzoni ha scritto:

Ah, yes, I've noticed it, but I thought it was "more right" using autogen.sh as in gpm package.
So I follow ALSA package model.

Thank you.

Giulio

Patches are now part of sources.

Delete patches.

---
 .../minicom/0001-musl-libc-compile-fix.patch  | 28 ----------
 ...-change-maxnamlen-to-posixs-name-max.patch | 51 -------------------
 2 files changed, 79 deletions(-)
 delete mode 100644 package/minicom/0001-musl-libc-compile-fix.patch
 delete mode 100644 package/minicom/0002-change-maxnamlen-to-posixs-name-max.patch

deleted file mode 100644
index 94c0da06a5..0000000000
@@ -1,28 +0,0 @@
-# HG changeset patch
-# User Adam Lackorzynski <[email protected]>
-# Date 1432407431 -7200
-#      Sat May 23 20:57:11 2015 +0200
-# Node ID 8bf81e6a8e60e37c2dec9577a1be6357b8d56165
-# Parent  8c784c80c17aa7a346d982db23bca5df67302cda
-musl-libc compile fix
-
-By Felix Janda <[email protected]>:
-VC_MUSIC is enabled by default on linux and in dial.c the
-necessary header are only included for glibc. (The wrong conditional
-include has likely been introduced by the 2003-03-30 GNU/Hurd patch.)
-
-Signed-off-by: Bernd Kuhls <[email protected]>
-[Taken from upstream Mercurial commit 8bf81e6a8e60]
-
-diff -r 8c784c80c17a -r 8bf81e6a8e60 src/dial.c
---- a/src/dial.c	Sat May 23 20:56:29 2015 +0200
-+++ b/src/dial.c	Sat May 23 20:57:11 2015 +0200
-@@ -39,7 +39,7 @@
- #include "intl.h"
- 
- #ifdef VC_MUSIC
--#  if defined(__GLIBC__)
-+#  if defined(__linux__) || defined(__GLIBC__)
- #    include <sys/ioctl.h>
- #    include <sys/kd.h>
- #    include <sys/time.h>

2.17.1

Minicom is now under git repo, not tar.gz anymore.

With git repo don't need hash file.

---
 package/minicom/minicom.hash | 2 --
 1 file changed, 2 deletions(-)
 delete mode 100644 package/minicom/minicom.hash

deleted file mode 100644
index ca3087160b..0000000000
@@ -1,2 +0,0 @@
-# Locally calculated
-sha256 532f836b7a677eb0cb1dca8d70302b73729c3d30df26d58368d712e5cca041f1  minicom-2.7.1.tar.gz

2.17.1

Hello,

All these patches should be just one, otherwise your series is not bisectable.

Best regards,

Thomas

Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com

Hello,

Il 01/08/2018 16:47, Thomas Petazzoni ha scritto:

Ah ok, I thought it was easier break all patches. So basically the commit log will be the cover letter, right? And btw, is commit log of cover letter acceptable? I don't know if it's too long or messy.
Please let me know,
thank you very much

Giulio

Hello,

It's not about being "easy". It's about separating logical changes, one per patch.

Here the problem of your patch series is that if I apply just PATCH 1/4, minicom no longer builds, because it doesn't do the autoreconf tweak, the patches won't apply, and the hash file doesn't match. This is why I said your series is "not bisectable": it doesn't build/work between each patch of the series.

In your situation, the change to use this different download location for Minicom is the change, and as part of the change, it is necessary to remove patches, tweak for autoreconf and delete the hash file. This is all one change.

Yes.

There is no "commit log" in the cover letter. The commit logs are in the patches themselves. The cover letter contains some introduction text, and is only necessary for patch series with multiple patches, for which additional explanations about what the whole series is doing.

Best regards,

Thomas

Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com

Hello,

Il 01/08/2018 19:00, Thomas Petazzoni ha scritto:

All clear now, thanks.

Yes, sorry i've used wrong words.

Thank you very much for explanation.
I follow with new patch.

Giulio