Gambit-list Septembre 2007

gambit-list@iro.umontreal.ca

25 participants
37 discussions

Edge triggered I/O with libevent
by Adam Langley 04 Avr '13

04 Avr '13

Gambit uses the select call on POSIX systems to multiplex IO operations. This is fine for small numbers of file descriptors, but falls apart quickly with large numbers[1]. libevent[2] is a wrapper library around a number of modern IO multiplexing systems (epoll on Linux, kqueue on the BSDs as well as poll and select). Attached is a patch which converts Gambit to using edge triggered I/O with epoll and libevent. This also requires a patch (attached) of libevent since the standard distribution doesn't include edge triggering support and the upstream appears to be on holiday. (epoll is required because I don't have a BSD box on which to develop edge triggering support for kqueue) This patch (for Gambit) preprocesses away to nothing unless --enable-libevent is passed to the configure script. (Note that, since this patches the configure.ac script, you'll need to run autoconf to regenerate). I hope I've not broken anything, but at this stage I've only tested that the console and tcp ports work. This patch (in the future) should also allow people, who write programs which use socket calls directly, to integrate into the main event loop at some point. Testing is more than welcome and, if you find a form of IO which is broken, please email me. [1] http://monkey.org/~provos/libevent/libevent-benchmark2.jpg [2] http://monkey.org/~provos/libevent/ -- Adam Langley agl(a)imperialviolet.org http://www.imperialviolet.org (+44) (0)7906 332512 PGP: 9113 256A CC0F 71A6 4C84 5087 CDA5 52DF 2CB6 3D60

2 2

Please don't delete the .c file
by Bradley Lucier 06 Fév '08

06 Fév '08

Marc: I needed to debug a gsc-compiled file with gdb today, and it was fairly inconvenient that the .c file wasn't left around by compile- file in beta 21. I suggest that the .c file be left around for the various source tools that need it. Brad

3 10

Tab completion
by Marc Feeley 17 Oct '07

17 Oct '07

I have added tab-completion to the builtin input line editor. When you press tab, the editor will extract the name preceding the cursor and search the symbol table for the lexicagraphically following symbol which will replace the original name. Pressing tab again will find the next symbol, and so on until the set of acceptable symbols is completely enumerated. Just get the latest changeset from the Gambit repository and run "make". Marc

4 4

working with ___arg* variables in c interface
by James Long 14 Oct '07

14 Oct '07

In a c-lambda, what are the types of the argument variables ___arg1, ___arg2, and so on? It seems like they are statically resolved to the types I passed in to the c-lambda, and everything happens that I expect to happen: (define %%addOne (c-lambda (int) int "___result=___arg1+1;")) (display (%%addOne 10)) This correctly prints out 11, and other various operations on pointers seemed to work too. For example, if 'ctx' is a structure with a 'id' field, and 'ctxp' is defined as (pointer ctx), the following function works fine: (define %%useCtx (c-lambda (ctxp) int "___result=(*___arg1).id;")) Basically, my question is when do I need to use ___CAST on the argument variables? Is it something I should be doing most of the time for safety, to suppress warnings or for some other reason? Or is it something done in a special case I haven't encountered yet? I've seen it done in other gambit FFI's and I'm wondering if the other was just paranoid or something (and if I should be).

3 2

process-status, close-port, and zombie processes
by Lang Martin 03 Oct '07

03 Oct '07

I've been building a system that runs (without a tty, if that makes a difference) in the background on a Linux system, and which, in turn runs a series of processes. It's in a fairly normal shell script pipeline and exists for the purpose of multi-threading this one step of the pipeline. All seems to be well running interactively. However, when the process runs as a gsi-script, it blocks inside ##thread-sleep! in ##process- status. The final result is the script running at 100% of the CPU (polling the port, seemingly) with a dangling child process, which has already exitted (exat?) and is now a zombie. It seems as though once this happens, each remaining open-process port will block until the full timeout of process-status is reached. the most relevant functions are here: (define (throw-shell-error port timeout) (or (zero? (process-status port timeout 100)) (error "fg command failed")) (close-port port)) (define (bg cmd . arguments) (open-process (list path: cmd arguments: arguments))) (define (fg cmd . arguments) (let ((proc (open-process (list path: cmd arguments: arguments)))) (throw-shell-error proc *join-timeout*))) I am running these commands purely for side-effects, and not reading from or writing to their ports. The problem appears to have something to do with load -- I can run a large number of low impact commands, and a much smaller number of "convert" (from the ImageMagick tools) commands before the zombie processes begin to block my execution. So, as I'm investigating: am I using these commands in the fashion intended by the author? I believe that they are correct in the sense of indicating the programs I mean to run, but am I misusing the port metaphor in some fashion? Any help would be appreciated. For a bit I had each process represented by a gambit thread, but I'm currently just mapping over collections of ports. Gambit v4.0.0, Linux 2.6.17.3 running under gsi-script, same behavior with and without -:d. I haven't used other options to gsi-script.

2 2

Re: [gambit-list] Generalized namespaces into aliases
by atsmyles＠rcn.com 29 Sep '07

29 Sep '07

Normally in scheme you have one namespace like this: name -> <location> The location can contain either a macro or a value. The identifer-macro in R6RS is like an ordinary macro except the the input form is an identifer. In that case you would have: name -> <identifier-macro transformer> An alias works differently, It is more like a pointer to another name. It doesn't directly point to a location in memory. So for an alias like (aliases (alias name)) would create a structure like: alias -> name -> <identifer-macro transformer> This is also how namespaces work in gambit currently, except in this syntax you specify the mapping explicitly. So (namespace ("test#") hello) is equivalent to (aliases (hello test#hello)). Arthur >> The idea is that a user can create aliases that are resolved before >> looking up functions or macros. >This looks a lot like symbol macros in Common Lisp or R6RS. >How do the two systems compare? Brad

1 0

Generalized namespaces into aliases
by atsmyles＠rcn.com 29 Sep '07

29 Sep '07

I implemented a feature to create aliases for functions and macros. The idea is that a user can create aliases that are resolved before looking up functions or macros. So for example, you could write: (aliases (kar car) (kdr cdr)) And from then on you can use those names in addition to the original names. The functionality is similar to the namespace mechanism except you can also create arbitrary renames instead of just prefixed names. More details and patches are at: http://www.iro.umontreal.ca/~gambit/bugzilla/show_bug.cgi?id=10 Arthur

2 1

gsc naming conflict with ghostscript
by naruto canada 27 Sep '07

27 Sep '07

hi There is a naming conflict with gambit-c and ghostscript: gsc GPL Ghostscript 8.56 (2007-03-14) ...... Does gambit-c have plan to make it install like production system then development system?

2 1

Node constraints for Termite?
by andrew cooke 27 Sep '07

27 Sep '07

Are there constraints on the number of *active* nodes in Termite? The reason I ask is that while I can create many nodes on the same machine, I can only remote-spawn to one other node (I am using a dual core machine). For example, using the code included (which is a bit odd because I have pulled code out of libraries to make a simple demo), the following is ok (with a single node) : gsi test.scm ((1 a) (2 b) (3 c) (1 d)) ("acp-make-nodes" #<node #2 ip: "localhost" port: 30000> ok) ("acp-map spawn" #<node #2 ip: "localhost" port: 30000>) ("acp-map spawn" #<node #2 ip: "localhost" port: 30000>) ("acp-map spawn" #<node #2 ip: "localhost" port: 30000>) ("acp-map spawn" #<node #2 ip: "localhost" port: 30000>) ("acp-map recv" #<thread #3>) ("acp-map recv" #<thread #4>) ("acp-map recv" #<thread #5>) ("acp-map recv" #<thread #6>) (2 3 4 5) but with two nodes created i see: : gsi test.scm ((1 a) (2 b) (3 c) (1 d)) ("acp-make-nodes" #<node #2 ip: "localhost" port: 30000> ok) ("acp-make-nodes" #<node #3 ip: "localhost" port: 30001> ok) ("acp-map spawn" #<node #3 ip: "localhost" port: 30001>) ("acp-map spawn" #<node #2 ip: "localhost" port: 30000>) ------------- REPL is now in #<thread #4> ------------- *** INTERRUPTED IN ##wait-for-io! (where I hit Ctrl-C after it hung) Any help appreciated - the enclosed file should run the first case above. To see the second change the appropriate line to (nodes (acp-make-nodes (list "localhost") 30000 2 #f))) - ie the penultimate arg is the number of nodes. Thanks, Andrew

2 1

call-with-input-file memory use
by Nicholas Walton 27 Sep '07

27 Sep '07

Hi, Is there any reason that reading a medium size file (10.9M) containing a fairly convoluted numeric list structure, previously written from Gambit, should take for ever using (call-with-input-file ...) and use an inordinate amount of memory. Trying to load twenty of these files in sequence blows my VM to 3.5Gigabyte before my Mac bottles out. If I change the datafile into a scm file by enclosing the data with (define x '.....) then the file loads immediately. Nick Walton

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Gambit-list Septembre 2007