- Move header licensing from tri-license to MPL 2.0. MPL-compatible
other licensing has been retained where originally present.
- Remove individual superseded licensing terms.
- Remove patches, outdated readmes & incomplete patch summaries.
- Remove incomplete cairo release notes (only went up to 1.6.4 anyway).
- Rewrite COPYING to indicate the current state of the library in tree.
This avoids getting data synchronously on the main thread in an XHR
(which has been deprecated for a long time and _may_ actually be blocked
in our networking) and attempts to be more predictable by always firing
an update request for the dialog from the XHR request handlers.
This also adds a reftest for border radius on collapsed borders (should
be ignored according to the CSS3 standard). We didn't do this before,
except on internal elements.
list items.
This patch does the following things:
1. Creates nsDisplayTableBorderCollapse that draws all collapse border
of tables.
2. Stops the use of nsDisplayTableBorderBackground.
3. Lets column and column group frames generate display items.
4. When traversing the table, also traverses the column and column group
frames.
5. For each type of table frame (col group, col, row group, row and
cell), draws their own background.
This should do it for all the commits to files I changed, but while I'm in here I could probably go ahead and turn ALL the singular if defined statements into ifdef statements by using grep/find on the tree. On the other hand, perhaps we should do that as a separate issue so that this doesn't become a case of scope creep.
OpenIndiana didn't need this for some reason, but on Oracle Solaris, we need this to make sure we're using gsed (GNU sed) here. It's probably a safer bet anyway.
This fix is a bit ugly and may need to be changed later if we switch a new GCC version, but the fact is that we use an architecture-specific path for GCC libraries on Solaris, so knowing the right prefix for GCC would only help so much, because it would still need to decide between ${gccdir}/lib and ${gccdir}/lib/amd64. The MOZ_FIX_LINK_PATHS variable puts the search paths into the right order without the need for me to use elfedit on the binaries afterwards.
https://developer.mozilla.org/en-US/docs/Mozilla/Developer_guide/Build_Instructions/Compiling_32-bit_Firefox_on_a_Linux_64-bit_OS
Setting this up turned out to be easier than I thought it would be. All I had to do was apply these instructions in reverse and add the following to my .mozconfig file:
CC="gcc -m64"
CXX="g++ -m64"
AS="gas --64"
ac_add_options --target=x86_64-pc-solaris2.11
export PKG_CONFIG_PATH=/usr/lib/amd64/pkgconfig
ac_add_options --libdir=/usr/lib/amd64
ac_add_options --x-libraries=/usr/lib/amd64
Most of these changes were fairly trivial, just requiring me to make a few of the changes I made earlier conditional on a 32-bit build. The biggest challenge was figuring out why the JavaScript engine triggered a segfault everytime it tried to allocate memory. But this patch fixes it:
https://github.com/OpenIndiana/oi-userland/blob/oi/hipster/components/web/firefox/patches/patch-js_src_gc_Memory.cpp.patch
Turns out that Solaris on AMD64 handles memory management in a fairly unusual way with a segmented memory model, but it's not that different from what we see on other 64-bit processors. In fact, I saw a SPARC crash for a similar reason, and noticed that it looked just like mine except the numbers in the first segment were reversed. Having played around with hex editors before, I had a feeling I might be dealing with a little-endian version of a big-endian problem, but I didn't expect that knowledge to actually yield an easy solution.
https://bugzilla.mozilla.org/show_bug.cgi?id=577056https://www.oracle.com/technetwork/server-storage/solaris10/solaris-memory-135224.html
As far as I can tell, this was the last barrier to an AMD64 Solaris build of Pale Moon.
https://bugzilla.mozilla.org/show_bug.cgi?id=1397928
Was looking into that _POSIX_PATH_MAX/NAME_MAX issue earlier because it didn't make a lot of sense and I was thinking of other approaches besides char arrays, and I wanted to make sure it didn't cause problems after they did it. Turns out that one commit after this was added, Mozilla determined the code I was working on fixing to be dead code as of Firefox 58. I don't know if it's dead code in Pale Moon as well, but given that it compiles fine without it and I can't find any other references to szExeFile in the IPC code, that seems like a safe bet.
Besides, I determined config/pathsub.c already seems to do what this code looks like it's trying to do, and implements the solution of just defining NAME_MAX to 256 and having done with it that I nearly adopted after realizing that even OS/2 and BeOS, let alone Unix/Linux systems, all basically use that value and there's just disagreement on which system header to check for it.
82dd4adb0e/components/desktop/firefox/patches/firefox-49-npapi.patch
The first fix was something I found on Oracle's patchset and allowed me to use the last Flash Player compiled for Solaris, from all the way back in 2012. Still works with most Flash content.
The second is an evolution of what I had to do to get Interlink to compile. For Interlink, I basically had to copy the contents of any boolean values from confvars.sh into the empty moz.configure file, otherwise nothing would get configured. I decided to test whether Pale Moon had the same issue, and it turned out that it wasn't as bad as on Interlink, but it was still pure luck that the browser component built at all, because MOZ_PHOENIX and other important flags were apparently not being defined at all, hence why I couldn't get half the extensions to be compatible at first.
I don't know why this is the case, but apparently configure.in isn't able to import values from confvars.sh. old-configure.in seems immune to the problem that application-specific configure.in files were experiencing, though. confvars.sh itself seems to work fine with values that aren't passed along via configure.in, though. So it's the interface between those two files that is messed up.
https://bugzilla.mozilla.org/show_bug.cgi?id=1369061
OpenIndiana used a much messier fix for this, but this one was used by Mozilla and looks a lot cleaner. It shouldn't interfere with any other targets, but if it does, the messy version of this fix basically involves rewriting the Solaris version of atomicops_internals_solaris.h to use GCC compiler intrinsics for atomic operations directly. It absolutely works, but it's gross to look at.
3b246b0b38/components/web/firefox/patches/01-FF43.0b3_OpenSXCE_x86_x64.patch
Another fix may be possible by preventing config/gcc-hidden.h from being included, or possibly using well-placed GCC pragmas to solve the visibility issues.
https://www.tachytelic.net/2019/01/grep-recursively/
During testing, I tried simply replacing grep with ggrep, which was non-portable but worked fine. Using an environment variable with os.getenv set to 'g' also worked, but the problem with that approach is that you have to set it manually and some times the script will mess up if you don't explictly define it to an empty string for platforms that don't need it.
Setting TOOLCHAIN_PREFIX to 'g' seemed to solve all of my problems except this one, and it appears to be the only non-portable use of GNU grep in the whole tree. To understand what I tried to do here, let me present with you with two commands that yield the same output on my machine:
find . -name '*.xul' | xargs grep -E "#include"
ggrep -E -r "#include" --include="*.xul" .
I just tried to make the Python script follow the logic of those two commands, though I'm not sure how well I succeeded since I got no errors. I visualized everything this way:
ggrep -E -r "<strings>" --include="<filesuffixes>" <path>
find <path> -name '<filesuffixes>' | xargs grep -E "<strings>"
And arranged it all accordingly to the best of my ability, though I should point out that Python is not my strong suit.
https://bugzilla.mozilla.org/show_bug.cgi?id=1185424http://www.mindfruit.co.uk/2012/06/relocations-relocations.html
The libxul.so fix was implemented by Mozilla in Firefox 57 and personally recommended to me by an Oracle employee on the OpenIndiana mailing list. It can easily be made ifdef XP_SOLARIS, but it seems like the new way is considered a better solution overall by the original author of the code that had it use that null pointer hack to begin with.
I can't link where I found the fix for libffi because I came up with it myself while looking at the way sysv.S does things. Something clicked in my brain while reading that mindfruit link above, though, and gave me enough of a sense of what was going on to be able to fix libffi.
The libffi fix looks a bit hairy because of all the FDE_ENCODE statements, but if you examine the code closely, you should find that it does exactly what it did before on all platforms besides Solaris. I later discovered that people who originally ported Firefox to Solaris never figured this out during the Firefox 52 era and had to use GNU LD for linking libxul.so while using the Sun LD for the rest of the build to make it work. For whatever reason, it works for me now without the GNU LD trick.
https://bugzilla.mozilla.org/show_bug.cgi?id=1350355
u_int32_t is not an stdint.h type. Windows already requires this, Solaris needs it too. If someone has a nit with this approach, the alternatives include:
1. Just replacing every instance of u_int32_t with uint32_t.
2. Including <sys/types.h> for Solaris only, which does define this.
3. Changing the original ifdef to be WIN32 || XP_SOLARIS
But it really doesn't matter how you solve this problem, all of the approaches are functionally equivalent, and this one has been used in Firefox since version 55. As far as I can tell, all it does is apply a fix that was being done for Windows already to any platform that needs it.
https://bugzilla.mozilla.org/show_bug.cgi?id=1375467
I would ifdef this, but it's been in Firefox since version of 56, and Petr Sumbara's explanation as to why it's wrong in the first place is so detailed that it's pretty obvious the code wasn't technically doing things properly to begin with.
Basically, they tried to redefine a system function after including the header file that declares it, and it caused problems on Solaris because libc functions are imported into the C++ std namespace in a different way that also complies with standards. So the existing implementation is technically bad code on all platforms, the Solaris implementation just uncovered the lack of standards compliance in the Mozilla code.
https://bugzilla.mozilla.org/show_bug.cgi?id=137126612e63e91af
This sounds absurd, and the fact that we had to change a C++ comment to a C-style comment on account of this may seem hilarious and inconsequential, but this isapparently not a joke and leaving it as it is now may be a bad idea.
https://bugzilla.mozilla.org/show_bug.cgi?id=1345102
I assess this change to be low-risk for the following reasons:
1. It has been in Firefox since version 55 without issues.
2. The current behavior is not POSIX compliant, and is retained in the one instance where the new functionality causes issues.
3. It makes safer assumptions about implementation details than what we have now.
https://bugzilla.mozilla.org/show_bug.cgi?id=1351309https://bugzilla.mozilla.org/show_bug.cgi?id=1309157
I assess this change to be low-risk for the following reasons:
1. It has been in Firefox since version 55 without issues.
2. It's nearly identical to the fix for bug 1309157 which is already in our tree, so that one would be causing problems if this one were going to.
3. Linux, Windows, and BSD are known to have a hypot function in their math libraries.
4. Even if it does break something, it should only break a test and not critical functionality.
https://bugzilla.mozilla.org/show_bug.cgi?id=1352449
Mozilla patch that's been in the code since Firefox 55. Seems like there have been no ill effects from implementing it, and it would only increase the portability of the UXP code. All the Solaris Firefox repos I've seen implement some variation on the jsexn patch, and this seems to be the cleanest version of it.
I can add ifdefs if needed or there are performance concerns associated with this patch, but I get the impression this alignment backlog issue might affect a few platforms other than Solaris, though none were named. Otherwise I think they wouldn't have used "platforms that need it" in plural form or failed to ifdef it.
Current versions of libcubeb already have a Sun audio implementation, but in Firefox 52 and earlier, this was all they had. I'm not completely happy with this implementation because it has issues like video freezing if the soundcard isn't working, but I think fixing this or pulling in a newer libcubeb would be going too far for too little gain.
https://bugzilla.mozilla.org/show_bug.cgi?id=1364865
Solaris doesn't define NAME_MAX because if you read the current POSIX standard literally, no system that supports multiple file systems or networking should be defining it. It's a pedantic choice given that they USED to define NAME_MAX, but Solaris always did take POSIX compliance seriously, for better or worse.
This is mostly ifdefs, but as you can see, Solaris is actually a lot like Linux. They're both more SysV than BSD at core, and most of the differences have more to do with Solaris not using glibc than anything else.
I still need to audit a lot of these changes and understand why they're needed and what the alternative approaches are. After this patch, most of the core functionality needed to build Solaris is here.
https://bugzilla.mozilla.org/show_bug.cgi?id=535106https://bugzilla.mozilla.org/show_bug.cgi?id=1359841
Like many parts of the busted Solaris support, this one has its origins in the pre-Firefox days. Bug 535106, another Mozilla suite bug. It keeps coming up because the core issue is never addressed, the fact that nsTSubstring doesn't know how to handle pid_t. I think the explicit cast to int is a band-aid they use because they know if they touch that substring header file to make it handle pid_t, they'll probably be asked to fix all the other problems with it. I honestly think it just works by accident on other platforms because it's implicitly cast to signed or unsigned int, even though the POSIX standard says pid_t can be either long or int, and work as either a signed or unsigned integer. Whatever the case may be, it's handled better on Solaris now than it was.
Ironically enough, the main point of having pid_t rather than just having pids be int or something is to hide this little implementation detail so you can just use pid_t for the return type in portable code without having to worry about what it is on a specific platform. The unfortunate way Mozilla implemented string functions turns that on its head and makes the good things about pid_t into liabilities rather than assets.
https://bugzilla.mozilla.org/show_bug.cgi?id=1360571
Solaris uses ksh as the default shell, and furthermore bash doesn't seem to support the 'local' keyword feature when invoked as sh on OpenIndiana. We could just change the script to use bash (it is an option even on Solaris), but this fix is available and Mozilla has been using it since Firefox 55 with no issues on any other platforms. It was specfically done this way to avoid any need to change existing mozconfig files, so I feel confident saying the change is totally benign and if anything the way it is now is technically a POSIX compliance issue inherited from Mozilla that we'll hit if we ever try to compile this on any Unix platform where bash isn't sh.
