Re: [testsuite] gdb.c++/templates.exp

Re: [testsuite] gdb.c++/templates.exp

[Michael Elizabeth Chastain]

Hi Michal,

This part is not quite correct:

  \\(unsigned ?(long|)\\);

It would accept 'unsignedlong' which would be wrong.

Can you try:

  \\(unsigned( long|)\\);

Or:

  \\((unsigned|unsigned long)\\);

Also, I need to know how you tested it.  It doesn't need to be an
extensive test, just say whether you've run it at least once,
and on which platform.

Michael C

===

2003-07-09  Michal Ludvig  <mludvig@suse.cz>

	* testsuite/gdb.c++/templates.exp (test_ptype_of_templates):
	Recognize expansion of size_t to 'unsigned long', not only 
	to 'unsigned'.

Re: [testsuite] gdb.c++/templates.exp

[Daniel Berlin]

On Wednesday, July 9, 2003, at 11:05 AM, Michael Elizabeth Chastain 
wrote:

> Hi Michal,
>
> This part is not quite correct:
>
>   \\(unsigned ?(long|)\\);
>
> It would accept 'unsignedlong' which would be wrong.

It shouldn't, unless the regex isn't whitespace sensitive
>
> Can you try:
>
>   \\(unsigned( long|)\\);

Now *this* would accept unsignedlong, unless the regexps in expect are 
unlike any other system i've seen.
This is because the alternation would be between " long" and "", *not* 
"long" and "" (IE your regex would not be equivalent to 
<space>long|<space>

Re: [testsuite] gdb.c++/templates.exp

[Daniel Berlin]

On Wednesday, July 9, 2003, at 11:15 AM, Daniel Berlin wrote:

>
> On Wednesday, July 9, 2003, at 11:05 AM, Michael Elizabeth Chastain 
> wrote:
>
>> Hi Michal,
>>
>> This part is not quite correct:
>>
>>   \\(unsigned ?(long|)\\);
>>
>> It would accept 'unsignedlong' which would be wrong.
>
> It shouldn't, unless the regex isn't whitespace sensitive

Just to be clear, this regex should accept nothing, "long", "unsigned 
long", and "unsigned".
This may be *wrong*, but it certainly shouldn't accept "unsignedlong".
The ? should be applying to "unsigned ", not just the whitespace.

>>
>> Can you try:
>>
>>   \\(unsigned( long|)\\);
>
> Now *this* would accept unsignedlong, unless the regexps in expect are 
> unlike any other system i've seen.
> This is because the alternation would be between " long" and "", *not* 
> "long" and "" (IE your regex would not be equivalent to 
> <space>long|<space>
>
I was on crack here and read it wrong, time to take my medication.
:)

Re: [testsuite] gdb.c++/templates.exp

[Daniel Jacobowitz]

On Wed, Jul 09, 2003 at 11:20:05AM -0400, Daniel Berlin wrote:
> 
> On Wednesday, July 9, 2003, at 11:15 AM, Daniel Berlin wrote:
> 
> >
> >On Wednesday, July 9, 2003, at 11:05 AM, Michael Elizabeth Chastain 
> >wrote:
> >
> >>Hi Michal,
> >>
> >>This part is not quite correct:
> >>
> >>  \\(unsigned ?(long|)\\);
> >>
> >>It would accept 'unsignedlong' which would be wrong.
> >
> >It shouldn't, unless the regex isn't whitespace sensitive
> 
> Just to be clear, this regex should accept nothing, "long", "unsigned 
> long", and "unsigned".
> This may be *wrong*, but it certainly shouldn't accept "unsignedlong".
> The ? should be applying to "unsigned ", not just the whitespace.

I have never encountered a regex engine which worked that way.  Unless
you parenthesize the ? will only apply to the previous character.  At
least perl, TCL, and egrep are (for once!) consistent here.  So is
procmail  :P

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [testsuite] gdb.c++/templates.exp

[Andreas Schwab]

Daniel Berlin <dberlin@dberlin.org> writes:

|> On Wednesday, July 9, 2003, at 11:15 AM, Daniel Berlin wrote:
|> 
|> >
|> > On Wednesday, July 9, 2003, at 11:05 AM, Michael Elizabeth Chastain
|> > wrote:
|> >
|> >> Hi Michal,
|> >>
|> >> This part is not quite correct:
|> >>
|> >>   \\(unsigned ?(long|)\\);
|> >>
|> >> It would accept 'unsignedlong' which would be wrong.
|> >
|> > It shouldn't, unless the regex isn't whitespace sensitive
|> 
|> Just to be clear, this regex should accept nothing, "long", "unsigned
|> long", and "unsigned".
|> This may be *wrong*, but it certainly shouldn't accept "unsignedlong".
|> The ? should be applying to "unsigned ", not just the whitespace.

It surely doesn't.  Repetition operators only apply to the smallest
preceding group.

The fixed prefix is "(unsigned", then comes an optional " ", then an
optional "long", then the fixed postfix ");".  All in all it matches
exactly four strings:

"(unsigned);"
"(unsigned );"
"(unsignedlong);"
"(unsigned long);"

Only the first and the fourth are desired.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [testsuite] gdb.c++/templates.exp

[Michal Ludvig]

[-- Attachment #1: Type: text/plain, Size: 409 bytes --]

Michael Elizabeth Chastain told me that:
> Can you try:
> 
>   \\(unsigned( long|)\\);

OK.

> Also, I need to know how you tested it.  It doesn't need to be an
> extensive test, just say whether you've run it at least once,
> and on which platform.

I have run the testsuite on i386 and amd64 and it worked.

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

[-- Attachment #2: testsuite-c++-templates-2.diff --]
[-- Type: text/plain, Size: 3033 bytes --]

2003-07-09  Michal Ludvig  <mludvig@suse.cz>

	* testsuite/gdb.c++/templates.exp (test_ptype_of_templates):
	Recognize expansion of size_t to 'unsigned long', not only 
	to 'unsigned'.

Index: testsuite/gdb.c++/templates.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.c++/templates.exp,v
retrieving revision 1.18
diff -u -p -r1.18 templates.exp
--- testsuite/gdb.c++/templates.exp	4 Mar 2003 17:02:20 -0000	1.18
+++ testsuite/gdb.c++/templates.exp	10 Jul 2003 06:55:05 -0000
@@ -61,7 +61,7 @@ proc test_ptype_of_templates {} {
 	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;((${ws}T5<int> & operator=\\(T5<int> const ?&\\);)|(${ws}T5\\(int\\);)|(${ws}T5\\((T5<int> const|const T5<int>) ?&\\);)|(${ws}~T5\\((void|)\\);)|(${ws}static void \\* operator new\\(unsigned( int| long)?\\);)|(${ws}static void operator delete\\(void ?\\*\\);)|(${ws}int value\\((void|)\\);))*${ws}\}\r\n$gdb_prompt $" {
 	    pass "ptype T5<int> (obsolescent gcc or gdb)"
 	}
-	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
+	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned( long|)\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
 	    # This also triggers gdb/1113...
 	    kfail "gdb/1111" "ptype T5<int>"
 	}
@@ -87,7 +87,7 @@ proc test_ptype_of_templates {} {
 	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;((${ws}T5<int> & operator=\\(T5<int> const ?&\\);)|(${ws}T5\\(int\\);)|(${ws}T5\\(T5<int> const ?&\\);)|(${ws}~T5\\((void|)\\);)|(${ws}static void \\* operator new\\(unsigned( int| long)?\\);)|(${ws}static void operator delete\\(void ?\\*\\);)|(${ws}int value\\((void|)\\);))*${ws}\}\r\n$gdb_prompt $" {
 	    pass "ptype t5i (obsolescent gcc or gdb)"
 	}
-	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
+	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned( long|)\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
 	    # This also triggers gdb/1113...
 	    kfail "gdb/1111" "ptype T5<int>"
 	}

Re: [testsuite] gdb.c++/templates.exp

[Michal Ludvig]

Michael Elizabeth Chastain told me that:
> Beautiful!  This patch is approved for both mainline and branch.

Thanks, committed.

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

Re: [testsuite] gdb.c++/templates.exp

[Michael Elizabeth Chastain]

Beautiful!  This patch is approved for both mainline and branch.

Michael C

===

2003-07-09  Michal Ludvig  <mludvig@suse.cz>

	* testsuite/gdb.c++/templates.exp (test_ptype_of_templates):
	Recognize expansion of size_t to 'unsigned long', not only 
	to 'unsigned'.

Re: [testsuite] gdb.c++/templates.exp

[Daniel Berlin]

On Wednesday, July 9, 2003, at 11:50 AM, Michael Elizabeth Chastain 
wrote:

> Hi Daniel,
>
>>>   \\(unsigned ?(long|)\\);
>
>> Just to be clear, this regex should accept nothing, "long", "unsigned
>> long", and "unsigned".
>> This may be *wrong*, but it certainly shouldn't accept "unsignedlong".
>> The ? should be applying to "unsigned ", not just the whitespace.
>
> The ? applies to one atom before it.  An atom is just the single
> character ' ', not the whole string 'unsigned '.  So says "Tcl and the
> Tk Toolkit", section 9.3.
>
> It's been that way in Unix regular expressions since the days of 'ed'.
>
> What with the fifty different regular expression implementations out
> there, maybe one of them does behave that way.

Two, actually.

> Not TCL or egrep or perl
> or vim though.
>
Of course, I never bothered to use one of these, i was working with a 
C++ regex engine.

> Michael C

Re: [testsuite] gdb.c++/templates.exp

[Michael Elizabeth Chastain]

Hi Daniel,

>>   \\(unsigned ?(long|)\\);

> Just to be clear, this regex should accept nothing, "long", "unsigned 
> long", and "unsigned".
> This may be *wrong*, but it certainly shouldn't accept "unsignedlong".
> The ? should be applying to "unsigned ", not just the whitespace.

The ? applies to one atom before it.  An atom is just the single
character ' ', not the whole string 'unsigned '.  So says "Tcl and the
Tk Toolkit", section 9.3.

It's been that way in Unix regular expressions since the days of 'ed'.

What with the fifty different regular expression implementations out
there, maybe one of them does behave that way.  Not TCL or egrep or perl
or vim though.

Michael C

Re: [testsuite] gdb.c++/templates.exp

[Michael Elizabeth Chastain]

Hi Daniel,

mec>   \\(unsigned ?(long|)\\);
mec>
mec> It would accept 'unsignedlong' which would be wrong.

berlin> It shouldn't, unless the regex isn't whitespace sensitive

I think you have it backwards:

  PATTERN       TEXT
  'unsigned' -> 'unsigned'
  ' ?'       -> ''
  '(long|)'  -> 'long'

Matches 'unsignedlong'.

Try it with egrep:

  $ egrep 'unsigned ?(long|)'
  unsignedlong
  unsignedlong

Try it with tclsh:

  $ tclsh
  % regexp "unsigned ?(long|)" unsignedlong
  1

mec> Can you try:
mec>
mec>   \\(unsigned( long|)\\);

berlin> Now *this* would accept unsignedlong, unless the regexps in expect are 
berlin> unlike any other system i've seen.
berlin> 
berlin> This is because the alternation would be between " long" and "", *not* 
berlin> "long" and "" (IE your regex would not be equivalent to 
berlin> <space>long|<space>

This accepts either '(unsigned);' or '(unsigned long);'.
It does not accept '(unsignedlong);'.  Try it with egrep or tclsh.
I did.

  $ tclsh
  % regexp "\\(unsigned( long|)\\)" "(unsigned)"
  1
  % regexp "\\(unsigned( long|)\\)" "(unsigned long)"
  1
  % regexp "\\(unsigned( long|)\\)" "(unsignedlong)"
  0

I agree that the alternation is between " long" and "".
These two regular expressions are equivalent:

  \\(unsigned( long|)\\);
  \\((unsigned long|unsigned)\\);

Michael C

[testsuite] gdb.c++/templates.exp

[Michal Ludvig]

[-- Attachment #1: Type: text/plain, Size: 211 bytes --]

Hi,
this is probably an obvious fix, but anyway. Someone please approve it 
for both mainline and 6.0. Thanks!

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* (+420) 296.545.373 * http://www.suse.cz

[-- Attachment #2: testsuite-c++-templates-1.diff --]
[-- Type: text/plain, Size: 3034 bytes --]

2003-07-09  Michal Ludvig  <mludvig@suse.cz>

	* testsuite/gdb.c++/templates.exp (test_ptype_of_templates):
	Recognize expansion of size_t to 'unsigned long', not only 
	to 'unsigned'.

Index: testsuite/gdb.c++/templates.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.c++/templates.exp,v
retrieving revision 1.18
diff -u -p -r1.18 templates.exp
--- testsuite/gdb.c++/templates.exp	4 Mar 2003 17:02:20 -0000	1.18
+++ testsuite/gdb.c++/templates.exp	9 Jul 2003 13:23:36 -0000
@@ -61,7 +61,7 @@ proc test_ptype_of_templates {} {
 	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;((${ws}T5<int> & operator=\\(T5<int> const ?&\\);)|(${ws}T5\\(int\\);)|(${ws}T5\\((T5<int> const|const T5<int>) ?&\\);)|(${ws}~T5\\((void|)\\);)|(${ws}static void \\* operator new\\(unsigned( int| long)?\\);)|(${ws}static void operator delete\\(void ?\\*\\);)|(${ws}int value\\((void|)\\);))*${ws}\}\r\n$gdb_prompt $" {
 	    pass "ptype T5<int> (obsolescent gcc or gdb)"
 	}
-	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
+	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned ?(long|)\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
 	    # This also triggers gdb/1113...
 	    kfail "gdb/1111" "ptype T5<int>"
 	}
@@ -87,7 +87,7 @@ proc test_ptype_of_templates {} {
 	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;((${ws}T5<int> & operator=\\(T5<int> const ?&\\);)|(${ws}T5\\(int\\);)|(${ws}T5\\(T5<int> const ?&\\);)|(${ws}~T5\\((void|)\\);)|(${ws}static void \\* operator new\\(unsigned( int| long)?\\);)|(${ws}static void operator delete\\(void ?\\*\\);)|(${ws}int value\\((void|)\\);))*${ws}\}\r\n$gdb_prompt $" {
 	    pass "ptype t5i (obsolescent gcc or gdb)"
 	}
-	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
+	-re "type = class T5<int> \{${ws}public:${ws}static int X;${ws}int x;${ws}int val;${ws}void T5\\(int\\);${ws}void T5\\((T5<int> const|const T5<int>) ?&\\);${ws}~T5\\(int\\);${ws}static void \\* operator new\\(unsigned ?(long|)\\);${ws}static void operator delete\\(void ?\\*\\);${ws}int value\\((void|)\\);${ws}\}\r\n$gdb_prompt $" {
 	    # This also triggers gdb/1113...
 	    kfail "gdb/1111" "ptype T5<int>"
 	}