Failing tests when ECMA scripting enabled

[aelmahmoudy]

The following 85 tests failed when I build with ECMA scripting enabled:

 19/103 customEvent.html
 20/103 doctype.publicId.html
 21/103 doctype.systemId.html
 22/103 document.anchors.html
 23/103 document.baseURI.html
 24/103 document.body.childNodes.html
 25/103 document.body.children.html
 26/103 document.body.id.html
 27/103 document.characterSet.html
 28/103 document.currentScript.html
 29/103 document.doctype.html
 30/103 document.documentElement.html
 31/103 document.documentURI.html
 32/103 document.domain.html
 33/103 document.getElementsByClassName.html
 34/103 document.getElementsByClassName_2.html
 35/103 document.getElementsByTagName.html
 36/103 document.images.html
 37/103 document.links.html
 38/103 document.location.href.html
 39/103 document.new.html
 40/103 document.querySelectorAll.html
 41/103 document.querySelector.html
 42/103 domparser.html
 43/103 element.appendChild.html
 44/103 element.attributes.html
 45/103 element.checked.html
 46/103 element.childElementCount.html
 47/103 element.classList.html
 48/103 element.className.html
 49/103 element.cloneNode.html
 50/103 element.closest.html
 51/103 element.contains.html
 52/103 element.dataset.html
 53/103 element.dir.html
 54/103 element.eventListener.html
 55/103 element.firstChild.html
 56/103 element.firstElementChild.html
 57/103 element.getAttribute.html
 58/103 element.getAttributeNode.html
 59/103 element.getElementsByTagName.html
 60/103 element.hasAttribute.html
 61/103 element.hasAttributes.html
 62/103 element.hasChildNodes.html
 63/103 element.href.html
 64/103 element.id.html
 65/103 element.innerHTML.html
 66/103 element.insertBefore.html
 67/103 element.isEqualNode.html
 68/103 element.isSameNode.html
 69/103 element.lang.html
 70/103 element.lastChild.html
 71/103 element.lastElementChild.html
 72/103 element.matches.html
 73/103 element.nextElementSibling.html
 74/103 element.nextSibling.html
 75/103 element.nodeName.html
 76/103 element.nodeType.html
 77/103 element.nodeValue.html
 78/103 element.outerHTML.html
 79/103 element.parentElement.html
 80/103 element.parentNode.html
 81/103 element.previousElementSibling.html
 82/103 element.previousSibling.html
 83/103 element.querySelectorAll.html
 84/103 element.querySelector.html
 85/103 element.remove.html
 86/103 element.setAttribute.html
 87/103 element.textContent.html
 88/103 eventListener.html
 89/103 fetch.html
 90/103 fragment.appendChild.html
 91/103 fragment.firstChild.html
 92/103 fragment.getElementsByTagName.html
 93/103 fragment.hasChildNodes.html
 94/103 fragment.new.html
 95/103 image.new.html
 96/103 keyboardEvent.html
 97/103 navigator.appCodeName.html
 98/103 node.nodeType.html
 99/103 nodelist.forEach.html
100/103 style.backgroundClip.html
101/103 window.setTimeout.html
102/103 URL.html
103/103 URLSearchParams.html

For a few tests, the stderr says Segmentation fault (core dumped), for most of the tests, stderr says -eval:1: unknown option

Full build log: https://salsa.debian.org/aelmahmoudy/elinks/-/jobs/6891196/raw

[rkd77]

Sorry, it is broken. ecmascript tests require either spidermonkey=true or quickjs=true or mujs=true .
I did not anticipate case that someone wants sm-scripting=true with spidermonkey=false (default).
In 1f2929f possibility to set sm-scripting=true without spidermonkey=true cause error.

[aelmahmoudy]

Actually it failed with mujs=true too

[aelmahmoudy]

tests failed with both ecmascript=true & sm-scripting=true Build log: https://salsa.debian.org/aelmahmoudy/elinks/-/jobs/6894664/raw

[aelmahmoudy]

Build log when it failed with mujs=true: https://salsa.debian.org/aelmahmoudy/elinks/-/jobs/6891489/raw

[rkd77]

I don't know. In container with debian trixie, I cloned repos mentioned in logs, built and installed required packages.
dpkg-buildpackage and tests passed, so how to reproduce it?

[rkd77]

From branch debian/javascript.

[aelmahmoudy]

I build on debian unstable, trixie is debian testing. Please send me the build log I am using debian/javascript branch indeed

[rkd77]

elinks.log.gz
../b.sh:

#!/bin/bash
dpkg-query --list
git reset --hard
rm -rf .pc
rm -f src/js/fetch.h
dpkg-buildpackage

[aelmahmoudy]

`dpkg-query --list` won't show the package versions , please use `dpkg -l`

[rkd77]

I cannot see difference.

[aelmahmoudy]

`dpkg -l` show pkg versions

[rkd77]

@aelmahmoudy, if you copy and paste command to direcory with tests test/js/assert, then error messages ought to be more verbose. Eg.:
cd test/js/assert

UBSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 MSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 MESON_TEST_ITERATION=1 ASAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1 MALLOC_PERTURB_=135 /builds/aelmahmoudy/elinks/debian/output/source_dir/obj-x86_64-linux-gnu/src/elinks --test 1 --eval 'set ecmascript.enable = 1' --eval 'set ecmascript.allow_xhr_file = 1' --eval 'set ecmascript.enable_console_log = 1' --eval 'set ui.sessions.fork_on_start = 0' --no-connect 1 doctype.publicId.html

[aelmahmoudy]

I tested on my system, and it didn't fail. Note though that I changed the command a bit to run elinks that is installed on my system, I got 2 binaries, one built against MuJS and the other built against QuickJS, both passed. Yet on Launchpad builder the tests fails with this error: ``` 'unknown': unknown terminal type. ``` This build was on Ubuntu focal

[aelmahmoudy]

I ran env on the builder before running the tests it showed that TERM=unknown

[rkd77]

I guess elinks requires a terminal. I'll make these tests optional, another bool option to the meson.

[aelmahmoudy]

I set TERM=dumb and the tests passed ! Maybe that would be better than disabling the JS tests

[aelmahmoudy]

By the way, I tried running this test interactively, by opening doctype.publicId.html, yet clicking the button does nothing

[rkd77]

I copied content from other files, but did not pay attention to it. Only assertions are meaningful there.

[aelmahmoudy]

I see. I tried elinks with one of either MuJS & QuickJS. I tried on sites like GitHub, but I don't see a difference in experience, there are still "Harmless buttons" for example.

[aelmahmoudy]

3 tests fail with MuJS engine:

=================================== 88/103 ===================================
test:         document.getElementsByClassName_2.html
start time:   05:22:30
duration:     30.04s
result:       exit status 2
command:      MSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 UBSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 MESON_TEST_ITERATION=1 ASAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1 TERM=dumb MALLOC_PERTURB_=15 /builds/aelmahmoudy/elinks/debian/output/source_dir/obj-x86_64-linux-gnu/src/elinks --test 1 --eval 'set ecmascript.enable = 1' --eval 'set ecmascript.allow_xhr_file = 1' --eval 'set ecmascript.enable_console_log = 1' --eval 'set ui.sessions.fork_on_start = 0' --no-connect 1 document.getElementsByClassName_2.html
----------------------------------- stderr -----------------------------------
SyntaxError: [string]:4: unexpected token in expression: '.'
==============================================================================
================================== 102/103 ===================================
test:         fetch.html
start time:   05:22:34
duration:     30.02s
result:       exit status 2
command:      MSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 UBSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 MESON_TEST_ITERATION=1 ASAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1 TERM=dumb MALLOC_PERTURB_=183 /builds/aelmahmoudy/elinks/debian/output/source_dir/obj-x86_64-linux-gnu/src/elinks --test 1 --eval 'set ecmascript.enable = 1' --eval 'set ecmascript.allow_xhr_file = 1' --eval 'set ecmascript.enable_console_log = 1' --eval 'set ui.sessions.fork_on_start = 0' --no-connect 1 fetch.html
----------------------------------- stderr -----------------------------------
SyntaxError: [string]:1: unexpected token: 'function' (expected ';')
==============================================================================

================================== 103/103 ===================================
test:         URLSearchParams.html
start time:   05:23:01
duration:     30.04s
result:       exit status 2
command:      MSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 UBSAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1:print_stacktrace=1 MESON_TEST_ITERATION=1 ASAN_OPTIONS=halt_on_error=1:abort_on_error=1:print_summary=1 MALLOC_PERTURB_=8 TERM=dumb /builds/aelmahmoudy/elinks/debian/output/source_dir/obj-x86_64-linux-gnu/src/elinks --test 1 --eval 'set ecmascript.enable = 1' --eval 'set ecmascript.allow_xhr_file = 1' --eval 'set ecmascript.enable_console_log = 1' --eval 'set ui.sessions.fork_on_start = 0' --no-connect 1 URLSearchParams.html
----------------------------------- stderr -----------------------------------
URLSearchParams.html
ReferenceError: 'URLSearchParams' is not defined
==============================================================================


Summary of Failures:

 88/103 document.getElementsByClassName_2.html TIMEOUT        30.04s   exit status 2
102/103 fetch.html                             TIMEOUT        30.02s   exit status 2
103/103 URLSearchParams.html                   TIMEOUT        30.04s   exit status 2

Ok:                 95
Expected Fail:      5
Fail:               0
Unexpected Pass:    0
Skipped:            0
Timeout:            3

Full build log: https://salsa.debian.org/aelmahmoudy/elinks/-/jobs/6917043/raw

I didn't pull xxd for that build, since MuJS support didn't require it, could this be the reason ?

[rkd77]

No, mujs is for fans of MuJS. I don't know how to implement these features: (Associative arrays, Promise and ...).

[aelmahmoudy]

Btw, commit 4b979b4 fixed the failing JS tests issue (not the MuJS specific ones)

[rkd77]

MuJS is waiting for someone to fix it. For now, do not package MuJS flavor.
Not too many sites are "usable" with elinks's js implementation, but a year ago was the same, so no big regression.