gcl-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] Re: /tmp still ?


From: Camm Maguire
Subject: [Gcl-devel] Re: /tmp still ?
Date: 18 Jun 2006 21:01:36 -0400
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Greetings!  OK, tempfile has been tested a little, and is now in
do-recompile, but not yet used inside the compiler when executing
e.g. (compile 'foo).  Have to think about the latter a bit, since we
need to keep the gazonk lsp and object files around after the compile
in some circumstances (e.g. the build of pcl).  The way the
'si::with-temp-file macro works now is to close and delete the
temporary file outside of its scope.

Internally, the mkstemp call is used -- this guarantees at the kernel
level (I believe) that the returned path was opened as a new file
atomically.  As this does not allow for pathnames with extensions, we
then rename the new handle to the requested extension without closing
the handle.  This is guaranteed to be atomic too, but what I did not
know was that if some other process had the destination path
previously opened, they still read the old data after the rename has
replaced the contents for new opens -- this is what we want.

Here is sample syntax:

(with-temp-file 
        (s pn) ("/tmp/foo" "lsp") 
        (prin1 "foo" s)
        (with-open-file (q pn) 
                (read q)))

Hope this works for everybody.

Take care,

Robert Boyer <address@hidden> writes:

> Look very much forward to your new tmp work, and all the
> other wonderful things you are up to.  So many.  Utterly
> amazing.
> 
> Please feel free to utterly ignore this message.  It has
> zero priority except to the extent you wish to assign it
> one.  I fear that I may be more trouble to you than I am
> worth with such reports.
> 
> Bob
> 
> -------------------------------------------------------------------------------
> 
> I just now tried again to build a 2.7.0 big, static, ansi
> GCL on elgin, but it broke with almost the same error as
> last time, something about a file of yours on elgin's
> directory /tmp.  Text from the error:
> 
> Callee STORE-CIRCLE sigchange NIL to ((T T
>                                        (INTEGER -2147483648 2147483647))
>                                       T), recompiling WIDGET-FUNCTION
> ;; Compiling /tmp/recompile.lsp.
> Error in SYSTEM::DO-RECOMPILE [or a callee]: Cannot create the file 
> /tmp/recompile.data.
> 
> Fast links are on: do (si::use-fast-links nil) for debugging
> Broken at CONDITIONS:ERROR.  Type :H for Help.
>  1 (Continue) Retry opening file #P"/tmp/recompile.data".
>  2 Retry opening file #P"/tmp/recompile.data".
>  3 Retry compiling file "/tmp/recompile.lsp".
>  4 Retry compiling file "/tmp/recompile.lsp".
>  5 Retry loading file "../tkl.o".
>  6 Retry loading file "../tkl.o".
>  7 Return to top level.
> dbl:TK>>
> NIL
> dbl:TK>>make[1]: Leaving directory `/v/filer3/v0q027/gcl-2.7.0/gcl-tk'
> cd info && make
> -------------------------------------------------------------------------------
> 
> These build-from-scratch runs sure take a while -- a half an
> hour of cpu time per try.
> 
>   1607.196u 159.209s 33:30.25 87.8%   0+0k 0+0io 110pf+0w
> 
> -------------------------------------------------------------------------------
> 
> There is more than 1 megabyte of output to such a make
> clean/configure/build.  It's not so easy to figure out
> whether anything has gone wrong, much less what, has gone
> wrong.  I count over 600 lines that could or might be easily
> interpreted by someone as dumb as me as warnings or errors
> in some sense or another.
> 
> Cf.
> 
>    http://www.cs.utexas.edu/users/boyer/foo.text
> 
> 
> -------------------------------------------------------------------------------
> 
> Here's the /tmp directory, where the collision is apparently
> occurring:
> 
>   /tmp:
>   used 1225 available 1925216
>   drwxrwxrwt   6 root  root   61440 Jun 18 10:26 .
>   drwxr-xr-x  35 root  root    1024 Jun 18 04:51 ..
>   drwxrwxrwt   2 root  root    4096 Jun 16 02:09 .ICE-unix
>   -rw-r--r--   1 camm  guest     53 Jun 17 13:47 gazonk_12300_0.lsp
>   -rw-r--r--   1 camm  guest   1248 Jun 17 13:47 gazonk_12300_0.o
>   drwx------   3 hunt  prof    4096 Jun 12 12:24 gconfd-hunt
>   drwx------   2 root  root   16384 Mar 10 15:19 lost+found
>   drwx------   2 hunt  prof    4096 Jun 17 12:33 orbit-hunt
>   -rw-r--r--   1 camm  guest 242930 Jun 17 14:41 recompile.c
>   -rw-r--r--   1 camm  guest 120463 Jun 17 14:41 recompile.data
>   -rw-r--r--   1 camm  guest  17612 Jun 17 14:41 recompile.h
>   -rw-r--r--   1 boyer prof   42322 Jun 18 10:26 recompile.lsp
>   -rw-r--r--   1 camm  guest 502907 Jun 17 14:41 recompile.o
>   -rw-------   1 boyer prof  196034 Jun 16 08:23 zman8jHCLd
> 
> -------------------------------------------------------------------------------
> 
> Again, 0 priority for this message, except to the extent it
> seems possibly useful to you to work on it.
> 
> Bob
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

>  LocalWords:  pn lsp prin foo tmp ansi




reply via email to

[Prev in Thread] Current Thread [Next in Thread]