You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Martin Kraemer <Ma...@mch.sni.de> on 1998/01/15 22:00:26 UTC

[PATCH] Improve implementation of -c/-Cdirective reading

*** WARNING *** THIS PATCH BREAKS MOD_PERL *** WARNING ***

The recently added -C/-c directive handling has a few bugs I didn't
notice at the time they were added:

*   a "resource leak", or at least improper use of pcfg_openfile()
    without the matching cfg_closefile(). [cfg_closefile() added.]

*   A mis-use of the pcfg_openfile() interface which (as its name
    suggests) is supposed to open a file. Instead a NULL pointer was
    passed. In fact, the parms.config_file pointer was apparently simply
    assigned "to calm down the compiler". [using pcfg_open_custom() now.]

*   Instead of using srm_command_loop() to read from the "array based
    file", the lines were extracted from the table manually and passed
    to handle_command() individually (doesn't this break the nested
    commands?!?)    [using srm_command_loop() now]

*   In pcfg_openfile(), the (debug) logging code would log an undefined
    errno string and would not report on fopen() errors.
    [reports potential errors _after_ opening the file now.]

*   In pcfg_openfile(), an undefined file pointer would be entered into
    the configfile_t  structure and be returned to the caller.
    [deleted.]

*   Structured commands like
	httpd -C '<IfModule mod_speling.c>' \
	      -C 'CheckSpelling On' \
	      -C '</IfModule>' ....
    would not work because they were executed line-by-line. (Ehhmm..
    I may be wrong on that. Correct me.).  Now, the "memory file"
    looks just like a regular config file to the config line parser.

*   And finally, even devices would be considered valid input for config
    file reading (/dev/zero certainly being the one with the largest
    number of bytes in it).
    [allowing only /dev/null as a convenience, no other devices now.]


*** WARNING *** THIS PATCH BREAKS MOD_PERL *** WARNING ***

That is because I switched to using srm_command_loop(). But the perl
hook was coded into the explicit
    while (array has more lines) handle_command()
loop. I think it did NOT belong there!

Can one of the perl gurus please have a look at the proper place where
the "PerlModule" command could be checked? Is it in srm_command_loop()?


    Martin
-- 
| S I E M E N S |  <Ma...@mch.sni.de>  |      Siemens Nixdorf
| ------------- |   Voice: +49-89-636-46021     |  Informationssysteme AG
| N I X D O R F |   FAX:   +49-89-636-44994     |   81730 Munich, Germany
~~~~~~~~~~~~~~~~My opinions only, of course; pgp key available on request

Re: [PATCH] Improve implementation of -c/-Cdirective reading (take 2)

Posted by Doug MacEachern <do...@telebusiness.co.nz>.
Martin Kraemer wrote:

> On Fri, Jan 16, 1998 at 11:24:23AM +1300, Doug MacEachern wrote:
> > +1 on concept.   the patch makes one .rej files which I applied by
> > hand, but my sore eyes can't see what's wrong:
> > util.c: In function `pcfg_openfile':
> > util.c:715: parse error before `*'
> > util.c:726: `param' undeclared (first use this function)
>
> Could have been a gcc bug(I'm using a different compiler here). Anyway
> I undid the prototype beautification and #ifdef'ed the fstat code.
>
> Furthermore, I  added a missing error printf and sent the whole thing
> thru indent, so here's take 2.

Sweetness.   Works great, +1

-Doug


[PATCH] Improve implementation of -c/-Cdirective reading (take 2)

Posted by Martin Kraemer <Ma...@mch.sni.de>.
On Fri, Jan 16, 1998 at 11:24:23AM +1300, Doug MacEachern wrote:
> +1 on concept.   the patch makes one .rej files which I applied by
> hand, but my sore eyes can't see what's wrong:
> util.c: In function `pcfg_openfile':
> util.c:715: parse error before `*'
> util.c:726: `param' undeclared (first use this function)

Could have been a gcc bug(I'm using a different compiler here). Anyway
I undid the prototype beautification and #ifdef'ed the fstat code.

Furthermore, I  added a missing error printf and sent the whole thing
thru indent, so here's take 2.

> isokay, the #ifdef MOD_PERL isn't needed anymore.
Removed that too.

    Martin
-- 
| S I E M E N S |  <Ma...@mch.sni.de>  |      Siemens Nixdorf
| ------------- |   Voice: +49-89-636-46021     |  Informationssysteme AG
| N I X D O R F |   FAX:   +49-89-636-44994     |   81730 Munich, Germany
~~~~~~~~~~~~~~~~My opinions only, of course; pgp key available on request

Re: [PATCH] Improve implementation of -c/-Cdirective reading

Posted by Doug MacEachern <do...@telebusiness.co.nz>.
Martin Kraemer wrote:

> *** WARNING *** THIS PATCH BREAKS MOD_PERL *** WARNING ***
>
> The recently added -C/-c directive handling has a few bugs I didn't
> notice at the time they were added:

sheesh, I am cursed, I can't even apply your patch to fix all the bugs I
added ;-(+1 on concept.   the patch makes one .rej files which I applied by
hand, but my sore eyes can't see what's wrong:
util.c: In function `pcfg_openfile':
util.c:715: parse error before `*'
util.c:726: `param' undeclared (first use this function)
util.c:726: (Each undeclared identifier is reported only once
util.c:726: for each function it appears in.)
util.c:727: `descr' undeclared (first use this function)
util.c:728: `getch' undeclared (first use this function)
util.c:729: `getstr' undeclared (first use this function)
util.c:730: `close_func' undeclared (first use this function)

It's been a long week and late night, could you just send me your util.c so I
can test?

> That is because I switched to using srm_command_loop(). But the perl
> hook was coded into the explicit
>     while (array has more lines) handle_command()
> loop. I think it did NOT belong there!
>
> Can one of the perl gurus please have a look at the proper place where
> the "PerlModule" command could be checked? Is it in srm_command_loop()?

isokay, the #ifdef MOD_PERL isn't needed anymore.

-Doug


Re: [PATCH] Improve implementation of -c/-Cdirective reading

Posted by Martin Kraemer <Ma...@mch.sni.de>.
On Thu, Jan 15, 1998 at 10:00:26PM +0100, Martin Kraemer wrote:
> *   And finally, even devices would be considered valid input for config
>     file reading (/dev/zero certainly being the one with the largest
>     number of bytes in it).
>     [allowing only /dev/null as a convenience, no other devices now.]

...and for WIN32 and _EMX_ the test should probably check for names as
"con" or "con:", "ptn", "prn:" and the like instead of S_IFREG(), or is
there an S_IFREG() in MSVC which returns FALSE for them?
What other pseudo devices besides "nul", "nul:" and the ones mentioned
above are there in Windoze?

    Martin
-- 
| S I E M E N S |  <Ma...@mch.sni.de>  |      Siemens Nixdorf
| ------------- |   Voice: +49-89-636-46021     |  Informationssysteme AG
| N I X D O R F |   FAX:   +49-89-636-44994     |   81730 Munich, Germany
~~~~~~~~~~~~~~~~My opinions only, of course; pgp key available on request