From: Pali =?ISO-8859-1?Q?Roh=E1r?= Date: Wed, 14 Nov 2012 23:38:39 +0100 Subject: [U-Boot] [PATCH v2 2/4] New command bootmenu: ANSI terminal Boot Menu support In-Reply-To: <20121113082741.E0E6B2002E1@gemini.denx.de> References: <1338136729-3907-1-git-send-email-pali.rohar@gmail.com> <1351769953-13560-3-git-send-email-pali.rohar@gmail.com> <20121113082741.E0E6B2002E1@gemini.denx.de> Message-ID: <1542900.KzlGXe2rqe@pali> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Hi, I will fix all mentioned style problems. On Tuesday 13 November 2012 09:27:41 Wolfgang Denk wrote: > > + /* key up was pressed */ > > + if (key == 'A') > > + key = 1; > > + /* key down was pressed */ > > + else if (key == 'B') > > + key = 2; > > + /* other key was pressed */ > > + else > > + key = 0; > > + > > + } > > use switch() ? > > Can we please avoid using hard-coded magic constants like 'A', > 'B' etc. here? > And what to use? ANSI sequence for key up is \e[1A (down is B). > > > + while ((option = bootmenu_getoption(i))) { > > + > > + sep = strchr(option, '='); > > + if (!sep) > > + break; > > Is there any specific reason for inventing yet another data > format here? Can we not for example re-use what we already > have in the hwconfig command, and use common code for parsing > the format? > > Using a '=' as separator here is not a good idea, IMO. > So which char is better for separator? Bootmenu line has config "name=command" and I think '=' is the best char for separating key and value pair. > > + /* Add U-Boot console entry at the end */ > > + if (i < 100) { > > Magic constant 100 ? > -th menu entry is stored in env bootmenu_ (you can look to README for API). So I need to create string "bootmenu_" and call getenv. I think it is very bad idea to use malloc for very allocating string of size: strlen("bootmenu_") + log10(i) + 2. I'm using array of size strlen("bootmenu_") + 3 allocated at stack which is enught for number less then 100. And I think 100 menu entries in uboot is really a lot of. If you do not think I can increase number to 1000 or 10000 (which increase size of string +2)... > > +cleanup: > > + iter = menu->first; > > + while (iter) { > > + entry = iter->next; > > + free(iter->title); > > + free(iter->command); > > + free(iter); > > + iter = entry; > > + } > > + free(menu); > > Make this a separate function? > Same as function bootmenu_destroy, so I changed code to call bootmenu_destroy. > > > + /* If delay is 0 do not create menu, just run first entry > > */ > > + if (delay == 0) { > > + option = bootmenu_getoption(0); > > + if (!option) > > + return; > > + sep = strchr(option, '='); > > + if (!sep) > > + return; > > That would be an error condition, would it not? Should this > not raise some error message, then? > Ok, I added error message. > > +int menu_show(int bootdelay) > > +{ > > + bootmenu_show(bootdelay); > > + return -1; /* -1 - abort boot and run monitor code */ > > +} > > If this function never returns anything else, why do we need > the return value at all? > You can see that this function is not static. Reason is that menu_show is uboot API function which must return int. -1 means abort booting and run monitor code. > > +int do_bootmenu(cmd_tbl_t *cmdtp, int flag, int argc, char > > *const argv[]) +{ > > + char *delay_str = NULL; > > + int delay = 10; > > + > > + if (argc >= 2) > > + delay_str = argv[1]; > > + > > + if (!delay_str) > > + delay_str = getenv("bootmenu_delay"); > > + > > + if (delay_str) > > + delay = (int)simple_strtol(delay_str, NULL, 10); > > + > > + bootmenu_show(delay); > > + return 0; > > +} > > Umm... don't we handle any errors at all? > I do not understand what you mean. > > +First argument of bootmenu command override bootmenu_delay > > env > I cannot parse this. > First argument of bootmenu command is delay and override env bootmenu_delay. > > +If env bootmenu_delay or bootmenu arg is not specified, > > delay is 10 seconds > Why not using CONFIG_BOOTDELAY as default instead? > Ok. > > +If delay is 0, no entry will be shown on screen and first > > will be called +If delay is less then 0, no autoboot delay > > will be used > What will happen then? How is this different from delay==0 ? > If delay is less then 0, bootmenu will be shown and autoboot will be disabled. So there will not be any timer which automatically boot first entry in time $delay. > > +Boot Menu always add menu entry "U-Boot console" at end of > > all entries + > > +Example using: > > + > > + setenv bootmenu_0 Boot 1. kernel=bootm 0x82000000 # Set > > first menu entry + setenv bootmenu_1 Boot 2. kernel=bootm > > 0x83000000 # Set second menu entry + setenv bootmenu_2 > > Reset board=reset # Set third menu entry + > > setenv bootmenu_3 U-Boot boot order=boot # Set > > fourth menu entry + setenv bootmenu_4 # Empty string is > > end of all bootmenu entries + bootmenu 20 # Run > > bootmenu with autoboot delay 20s > You might give an example here what the resulting screen will > look like. > Ok, I can here text output. If you want to see screenshot from bootmenu rendered to framebuffer (in qemu) here is: http://atrey.karlin.mff.cuni.cz/~pali/u-boot-bootmenu.png > > > + #define CONFIG_BOOTDELAY 30 > > + #define CONFIG_AUTOBOOT_KEYED > > Do we really need CONFIG_AUTOBOOT_KEYED ? > Yes, because menu_show() call in common/main.c is called only of CONFIG_AUTOBOOT_KEYED is defined. You can look at this condition. -- Pali Roh?r pali.rohar at gmail.com -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 198 bytes Desc: This is a digitally signed message part. URL: