Date: Sun, 7 May 2000 19:18:57 +0930 (CST) From: Greg Lewis <glewis@trc.adelaide.edu.au> To: Fuyuhiko Maruyama <fuyuhik8@is.titech.ac.jp> Cc: glewis@trc.adelaide.edu.au, freebsd-java@freebsd.org Subject: Re: Alpha JDK 1.2.2 patchset 7 Message-ID: <200005070948.TAA51618@ares.trc.adelaide.edu.au> In-Reply-To: <551z3gqtah.wl@is.titech.ac.jp> from Fuyuhiko Maruyama at "May 6, 2000 11:57:58 am"
next in thread | previous in thread | raw e-mail | index | archive | help
Hi Fuyuhiko-san, An excellent set of reports and patches as usual :). Thanks! > 1. The _sigprocmask function (src/freebsd/hpi/green_threads/src/signals.c) > has never been called. > I think the _sigprocmask function is existing for wrapping syscall > by way of management the consistency of signal masks between each > green_threads, so it should be called for all unsafe (not well > handled or considered by VM's self) sigprocmask call, safe calls > may be written to call green_sigprocmask. Since patchset 7 and > older's are named _sigprocmask (with underscore!), it has never > called. It may be better to be renamed. I've gotten this reported independently in private email too. The patch from that email did things slightly differently, but it looks to me like the result is the same. > 2. A part of awt's internationalization doesn't work with non-Sun's > Motif. > src/freebsd/native/sun/awt/awt_util.c's awt_util_getIMStatusHeight > function may be used to calculates the Status Lines' height of > given Widget using *VENDOR PRIVATE AREAS*, and I feel the private > structure of Sun's Motif is completely different from others > including Software2Go's, Lesstif and (may be Metroworks's). > At this time, our best solution is always returning 0. In fact, I > think this is not so bad, because it may only needed for Solaris's > Motif integrated IM supports which doesn't exist in another Motif. > This is just the same solution of Blackdown's 1.2-pre-v2. > > It helps to run demo/applet/DitherTest. I tried merging in the version of awt_util_getIMStatusHeight from the Linux source code that Sun released, but appletviewer just dumps core when I try to run DitherTest :(. I think I will try returning zero as you say (at least while we are still using Lesstif). > 3. JIT compilers cannot handle NullPointerExceptions in JIT compiled > codes. > The function InitializeSignals() in > src/freebsd/javavm/runtime/signals_md.c installs some signal > handlers for JIT. In FreeBSD version, the handler for SIGSEGV is > commented out. It is needed not only for sunwjit but also OpenJIT > and shuJIT, FreeBSD version of tya seems to not use SIGSEGV but it > checks whether the pointer is null everytime (this costs too high!). Ok, I'll add this in. > 4. Incompatible type between int32_t(int, FreeBSD defined) and > jint(long, JDK defined) for x86. > This is not so serious problem, but javah outputs prototypes using > int32_t for Java's int when it works for old Native Method > Interface. In Solaris there is same problem and its solution is > defining int32_t long in src/solaris/javavm/include/typedefs_md.h. Right. We definitely need the uint32_t definition for consistency (I'll remove the one down lower for FreeBSD 2.2.x). I don't think we need the int8_t definition as well though, as this should match the FreeBSD definition. Why Sun have defined a 32 bit integer type that isn't 32 bits is a bit beyond me though... -- Greg Lewis glewis@trc.adelaide.edu.au Computing Officer +61 8 8303 5083 Teletraffic Research Centre To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200005070948.TAA51618>