AnnaH Report post Posted August 25 libxml2 2.12 is the only version on the system installed, so i think it uses correct one the code inside condition: #if LIBXML_VERSION >= 21200 xmlCtxtSetMaxAmplification (ctxt, 100); #endif is definitely ran 4900 #if LIBXML_VERSION >= 21200 4901 xmlCtxtSetMaxAmplification (ctxt, 100); 4902 printf("set new max amplification, i think?\n"); 4903 #endif ➜ eternallands ./el.linux.bin set new max amplification, i think? ➜ eternallands Testing with xmllint: ➜ eternallands xmllint --version xmllint: using libxml version 21200 compiled with: Threads Tree Output Push Reader Patterns Writer SAXv1 FTP HTTP DTDValid HTML C14N Catalog XPath XPointer XInclude Iconv ICU ISO8859X Unicode Regexps Automata Schemas Schematron Modules Debug Zlib ➜ eternallands xmllint --noent actor_defs/actor_defs.xml actor_defs/actor_defs.xml:143: parser error : Maximum entity amplification factor exceeded, see xmlCtxtSetMaxAmplification. &elf_female; ^ ➜ eternallands xmllint --noent --max-ampl 100 actor_defs/actor_defs.xml | wc -l 80618 (that's a lot of lines!) It looks like it reads the defs in ok with functon read_actor_defs, ok variable is = 1 all the way through the if block in read_actor_defs function: 4881 int read_actor_defs (const char *dir, const char *index) until getting assigned 0 by return val from parse_actor_defs function, at line: 4924 ok = parse_actor_defs (root); at second iteration of 'for' loop in parse_actor_defs function, first iteration: ok = 1, then ok = 0, wish i knew why, or how to test whats going on better Quote Share this post Link to post Share on other sites
AnnaH Report post Posted August 25 I narrowed it down to problem with sound in: 4476 int parse_actor_nodes(actor_types *act, const xmlNode *cfg, const xmlNode *defaults) i added printfs as this: 4546 } else if (!strcmp(name, "neck")) { 4547 ok &= parse_actor_neck(act, item, defaults); 4548 } else if (!strcmp(name, "sounds")) { 4549 #ifdef NEW_SOUND 4550 printf("neck ok: %i\n", ok); 4551 ok &= parse_actor_sounds(act, item->children); 4552 printf("sound ok: %i\n", ok); 4553 #endif //NEW_SOUND output when running el: ➜ eternallands ./el.linux.bin neck ok: 1 sound ok: 0 neck ok: 1 sound ok: 0 neck ok: 1 sound ok: 0 neck ok: 1 sound ok: 0 ... recompiled with NEW_SOUND commented out in make.defaults: 12 #FEATURES += NEW_SOUND # Enables extended sound effects system and game loads and can log in chars. found out that this is the culprit for me: 3620 if (!have_sound_config) return 0; inside: 3611 int parse_actor_sounds(actor_types *act, const xmlNode *cfg) 3612 { 3613 const xmlNode *item; 3614 char str[255]; 3615 int ok; 3616 int i; 3617 3618 if (cfg == NULL) return 0; 3619 if (!have_sound_config) return 0; // returns here, hence "ok &=" in caller gets zeroed 3620 Looks like sound system for EL is broken for me. Im using pulseaudio as my sound server, but EL for some reason wants to use pipewire. Which is installed on system as dependency of some packages but not used as sound server. Now to figure out how to make EL keep using pulseaudio. Quote Share this post Link to post Share on other sites
AnnaH Report post Posted August 25 (edited) Ok so having #disable_sound= 1 in el.ini will trigger the if statement diff --git a/actor_scripts.c b/actor_scripts.c index be4fd426..530a2f06 100755 --- a/actor_scripts.c +++ b/actor_scripts.c @@ -11,6 +11,7 @@ #include "draw_scene.h" #include "elconfig.h" #include "errors.h" +#include "elloggingwrapper.h" #include "gamewin.h" #include "hud_statsbar_window.h" #include "interface.h" @@ -4922,7 +4923,14 @@ void init_actor_defs() memset (actors_defs, 0, sizeof (actors_defs)); memset (attached_actors_defs, 0, sizeof (attached_actors_defs)); set_invert_v_coord(); - read_actor_defs ("actor_defs", "actor_defs.xml"); + if (read_actor_defs ("actor_defs", "actor_defs.xml") != 1) + { + char *message = "This is likely the libxml2 > v2.10 issue\nSee https://www.eternal-lands.com/forum/index.php?/topic/61996-cant-log-in-the-rulesxml-file-was-not-found\n"; + LOG_ERROR(message); + fprintf(stderr, message); + FATAL_ERROR_WINDOW("Failed to read actor defs. See Going to exit."); + exit(1); + } } from commit: commit d067d1548c8906dc1ef56d6b2cb9bd901612caf6 Now i need to figure out how to deal with this: [ALSOFT] (EE) Failed to connect PipeWire event context (errno: 112) but that's different topic Edited August 25 by AnnaH Quote Share this post Link to post Share on other sites
bluap Report post Posted August 25 (edited) Wow, that's an unintended consequence of adding the return value check for the read_actor_defs() call! I've had a quick look at what you have found and I think you found a long standing bug. If the statement: if (!have_sound_config) return 0; returns zero, the defs for sound are never loaded, even when you enable sound later. I would suggest we just remove the check completely which will fix part of the long standing bug and also fix the issue you have with return value check for the read_actor_defs() call failing too. Edit: Perhaps safer for now, leave the have_sound_config and return 1 to just prevent the error. I'll check some more later, but what do you think? Edited August 25 by bluap Quote Share this post Link to post Share on other sites
AnnaH Report post Posted August 25 (edited) Yep, I've seen your latest fix in repo, looks good! Thanks (also fixed my EL sound by just recompiling OpenAL without pipewire support, so that's good too) (also also, sorry A. for spamming your thread instead of using edit, in my defense it was 5am) Edited August 25 by AnnaH Quote Share this post Link to post Share on other sites
bluap Report post Posted August 25 Thanks for finding the problem with my additional checks, that could have lay hidden for ages. Glade you fixed your audio, we'll need to keep a look out for that. Quote Share this post Link to post Share on other sites
Erdie Report post Posted October 1 @bluapI am not very familiar with creating makefiles, therefore I ask for help. I was able to build an old version of libxml2 which did not have the problem we are discussing. I put it into a separate directory e. g. the EL build dir or even runtime dir. Of course el still uses the system wide libxml2 because of the systems preferences (even after rebuild). Can you tell me how to convince the build system to use the separate version in the current Dir of EL while modifying the "Makefile.linux". This should be possible, isn´t it? Thanks Erdie Quote Share this post Link to post Share on other sites
bluap Report post Posted October 3 Which distro are you using? On Arch for example, I have installed the old version and will hold the package until the updated version arrives. If you build your own package, most packages will support installing into the /usr/local tree which would be picked up by the build and at runtime, hopefully. If not you may have to remove the old package too. Quote Share this post Link to post Share on other sites
AnnaH Report post Posted October 3 (edited) If you feel like experimenting, you may try pointing build system to custom location of your libraries if they are not in default locations on system. Never done it with EL, but generally in other projects its done with flags to preprocessor and linker, e.g.: CPPFLAGS="-I /your/path/to/libxml2/include ${CPPFLAGS}" LDFLAGS="-L/your/path/to/libxml2/lib ${LDFLAGS}" and option -rpath for runtime linking with LDFLAGS="-Wl,-rpath,/your/path/to/libxml2/lib ${LDFLAGS}" edit: looks like el's Makefile doesn't use CPPFLAGS, so just prepend the "-I ..." option from CPPFLAGS to both CFLAGS and CXXFLAGS edit 2: since you are also on Gentoo, another option is to use version of libxml2 from upstream github by unmasking version "9999" with echo '=dev-libs/libxml2-9999 **' >> /etc/portage/package.accept_keywords/libxml2 and rebuild libxml2 emerge --oneshot --ask --verbose --quiet-build dev-libs/libxml2 which you can later update with emerge @live-rebuild target to emerge, because 9999 version are not picked up in normal emerge --update @world target. And when version 2.12 gets stabilized in portage tree, switch back by removing unmask from accept_keywords Edited October 7 by AnnaH Quote Share this post Link to post Share on other sites
Erdie Report post Posted October 6 On 10/3/2023 at 11:16 PM, bluap said: Which distro are you using? On Arch for example, I have installed the old version and will hold the package until the updated version arrives. If you build your own package, most packages will support installing into the /usr/local tree which would be picked up by the build and at runtime, hopefully. If not you may have to remove the old package too. I am using Gentoo. If your custom lib is loaded depends on the search priority where the system is looking for the libs. I can use /usr/local/lib but the problem is, that /usr/lib will be search before /usr/local/lib and if it is found, the version of /usr/local/lib will not be used. BTW: I masked the new version of libxml2 up to now and that the whole system uses the old version but this cannot be done forever because there are other apps using libxml2 and some of them might get broken once they are no more compatible with version 10.4.* Therefore I am looking for a better way e.g. convince the EL build system to look into the local dir ./. before looking into/usr/lib. But I do not exactly know how to archive that. Quote Share this post Link to post Share on other sites
Aislinn Report post Posted November 21 It's time! 2.12 released! Going to give it a whirl, fingers crossed. warning: libxml2: ignoring package upgrade (2.10.4-6 => 2.12.0-1) Result: Nope. Still not working. [actor_scripts.c:4931] Failed to read actor defs. See logs, going to exit. OK Edit 2: I recompiled and got this error: > named_colours.cpp: In member function ‘void ELGL_Colour::Colour_Container::load_xml()’: named_colours.cpp:103:28: error: ‘xmlReadFile’ was not declared in this scope 103 | if ((doc = xmlReadFile(file_name.c_str(), NULL, 0)) == NULL) | ^~~~~~~~~~~ make: *** [Makefile.linux:171: named_colours.o] Error 1 Quote Share this post Link to post Share on other sites
AnnaH Report post Posted November 21 (edited) 7 hours ago, Aislinn said: Edit 2: I recompiled and got this error: named_colours.cpp:103:28: error: ‘xmlReadFile’ was not declared in this scope 103 | if ((doc = xmlReadFile(file_name.c_str(), NULL, 0)) == NULL) | ^~~~~~~~~~~ diff --git a/asc.c b/asc.c index dc046479..79057202 100644 --- a/asc.c +++ b/asc.c @@ -5,6 +5,7 @@ #include <ctype.h> #include <iconv.h> #include <errno.h> +#include <libxml/encoding.h> #include "asc.h" #include "errors.h" #include "md5.h" diff --git a/named_colours.cpp b/named_colours.cpp index 6d46b4f9..65874370 100644 --- a/named_colours.cpp +++ b/named_colours.cpp @@ -15,6 +15,7 @@ #include <string> #include <cstring> #include <sstream> +#include <libxml/parser.h> #include "asc.h" #include "elloggingwrapper.h" you can patch with those two #includes for now Edited November 21 by AnnaH Quote Share this post Link to post Share on other sites
Aislinn Report post Posted November 26 Patch worked great, thank you! Quote Share this post Link to post Share on other sites
Erdie Report post Posted November 30 Now with the patch it compiles - does it mean the client is now also working with libxml2-2.12 ? Quote Share this post Link to post Share on other sites
