NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part2/18098

Text File | 1996-08-05 | 20.0 KB | 544 lines

Path: li.net!usenet
From: [e-mail] (Bob Silvern)
Newsgroups: comp.lang.c,comp.lang.c++
Subject: Borland should be ashamed.  My experience with BC5.0.
Date: Thu, 18 Apr 1996 21:42:03 GMT
Organization: Harmony Graphics
Message-ID: <4l6d1b$[e-mail]>
NNTP-Posting-Host: 199.173.74.139
X-Newsreader: Forte Free Agent 1.0.82

With the release of BC5.0, Borland has made it clear that they value
time-to-market far more than customer satisfaction.  Within 5 minutes of use I
found a considerable number of serious bugs, even with Patch 1 installed.
There is no way that Borland could have been ignorant of these problems.  I
have no doubt that they made the conscious decision to ship this product
prematurely in order not to lose market share to Microsoft and Symantec.

This is not the first time they have shipped a useless product.  Long time
Borland users will remember BC4.0 as another fiasco.  Although many of the
problems were fixed with later patches, the frustration endured by their
customers is not easily forgotten.  The short short-sightedness of Borland's
management will ultimately catch up with them.  I now regret giving them a
second chance by purchasing BC5.0.

On a slightly different subject, I would like to mention that Borland's much
touted bug reporting system, found at http://loki.borland.com/Cpp/BugSrch.htm,
is just another scam by Borland's marketing department.  While in theory this
would have been a very useful source of information, I believe that Borland has
made a conscious decision to list only minor bugs or those which have already
been fixed, leading would-be buyers to believe BC5.0 is a stable product.  I
have come to realize this after observing that *none* of the bugs reports
quoted below have appeared on Borland's bug reporting system, though most were
received by Borland nearly three weeks ago.

I suspect therefore, that many of us are submitting bug reports to Borland, in
many cases the same stupid, obvious bugs, over and over, which Borland knew
about from the start anyway, and which Borland refuses to publicly acknowledge.
I have included below some of the important bugs reports I've already submitted
to Borland, along with their quoted response (edited for brevity, especially
the auto-responder responses).

I invite other BC5.0 users to add to this thread any other bugs which are not
yet available on Borland's web site.  My bug reports follow:


*******************************************************************************************
My bug report #0, Borland bug report R001927
March 31, 1996
Subject: Viewing Globals:

    1) Typing more than one character makes some or all of the matching names
disappear.
    2) If browser is maximized, hitting backspace deletes characters from the
editor window instead of the browser.
    3) If browser is un-maximized, it becomes a blank gray window.

>April 3, 1996
>We were not able to reproduce number one or two.  If you
>have precise steps that we can use to recreate this behavior
>please submit a new bug report with the detailed steps to
>reproduce (i.e. Open example project Blazer, got to View|Globals
>etc...)
>Number three is a known problem and we are working towards
>fixing this issue.
>If you have further information regarding this issue, we would
>appreciate any additional assistance you could provide.  Please
>make sure to include the above reference number on any
>additional information that you provide.

April 4, 1996
The problem occurs with all projects on my system.  For example,
windows/examples/cmdlg/cmdlg.ide.
The steps to demonstrate problems 1 and 2 previously described are:
  1) Open an editor window containing cmdlg.c.
  2) Maximize the window and place the cursor at the end of a line of text.
Note the text left of the cursor because it will get deleted later.
  3) View, Globals.
  4) Place the cursor in the Search box. (Note that alt-h doesn't put the
cursor there as implied by the underlined h in "Search".  This is another bug
which I neglected to mention.)
  5) Type "w".
  6) 7 types and 1 function are displayed.  The 1 function is "int
WinMain(HINSTANCE__ *, HINSTANCE__ *,signed char *,int)".   So far, so good.
  7) Now type "i" so that the Search box reads "wi".  All 8 items are erased
from the display, including WinMain, which of course should still be visible.
This is problem #1.
  8) Now hit the backspace key a few times.  The characters in the search box
are not deleted.  Now look at the editor window.  The backspace key was
deleting the characters in the editor window, not the Search box.  This is
problem #2.  For 3 days I was unjustly accusing my cat of walking on the
keyboard as characters seemed to randomly disappear.  She has vowed (meowed)
never to buy another Borland product again.

Note: I have installed Turbo Assembler V5.0 after BC5.0.  Don't know if this is
significant.

>April 05, 1996
>Thank you for updating the information concerning your bug report.
> The new information has been accepted by the Techinical [sic] Support
> department for further investigation.
>
> Your reference number for this bug is still:   R002006

<Note: See my bug report #13.1>



*******************************************************************************************
My bug report #1, Borland bug report R001905
March 31, 1996
Subject: Bookmarks are not saved when exiting the IDE.

>April 03, 1996
>Comments and suggestions are important parts of product planning
>and development here at Borland.  We appreciate your input to help
>us plan and develop products that better meet the needs of you,
>our valued customer.




*******************************************************************************************
My bug report #2, Borland bug report R001906
March 31, 1996
Subject:  Editor Auto-save mode does not work.  If IDE crashes while running my
program, my edits are lost.


>April 01, 1996
>Thank you for submitting a bug report.  Your report has been accepted
>by the Techinical Support department for further investigation.
>
>Your reference number for this bug is:   R001906


*******************************************************************************************
My bug report #3, Borland bug report R001917
March 31, 1996
Subject:  Maximized editor window doesn't work right:

Start from the maximized message window, make the active line a message about
foo.c.  If you switch to an editor window of foo.c via the window menu or
Cntl-F6, you get a maximized editor window with the other windows on top of it.
Clicking cursor anywhere in another window maximizes that window.

>April 04, 1996
>After carefully reviewing your report, we have been able to
>reproduce the problem you described. This information will
>be forwarded to our development team for further processing.




*******************************************************************************************
My bug report #4, Borland bug report R001907
March 31, 1996
Subject:  Dialogs which contains BWCC controls are not displayed.  Their
procedures receive the WM_DESTROY and WM_NCDESTROY messages only.

>April 02, 1996
>I cannot reproduce the problem you describe using the
>options dialog box of the TTT example in OWL\GAMES.
>You may be running into a problem with BWCC
>registration.  Are you calling EnableBWCC()?  If so,
>then I would guess that you are running into a BWCC
>bug that is specific to 32-bits.  If so, download the file
>regdiag.zip from the OWL2 section of the Borland
>Download BBS at 408 431 5096.  This contains a
>workaround for this specific 32-bit case.

April 04, 1996
I have verifed the previously reported problem with a Borland example project
as described below:
  1) Open examples\windows\truetype\truetype.ide.
  2) Edit the About dialog box.
  3) Add a BWCC checkbox to the dialog box.
  4) Go to target expert and check the BWCC option.
  5) Make the project and run.
  6) Select File, About, and ...  nothing happens!
  7) Deleting the BWCC checkbox and adding a standard checkbox does not cause
this problem.

>April 05, 1996
>Thank you for updating the information concerning your bug report.
> The new information has been accepted by the Techinical  [sic] Support
> department for further investigation.
>
> Your reference number for this bug is still:   R002006

<Note: See My bug report #13.1>



*******************************************************************************************
My bug report #5, Borland bug report R001915
March 31, 1996
Subject:  CPU Window: Toggle Breakpoint speedbar button doesn't work.

>April 02, 1996
>We have carefully analyzed your bug report and determined that,
>while it is in fact a bug and has been entered in our bug database,
>there is a relatively simple workaround: the F5 key will toggle a
>breakpoint at the position of the current cursor location.




*******************************************************************************************
My bug report #6, Borland bug report R001916
March 31, 1996
Subject:  Changing a variable from inspect window, or Evaluate window, doesn't
always stick.

After single stepping a few times, variable is set back to original value.
Variable was in this case a long int parameter to the current function.

>April 03, 1996
>Based on the little information supplied I could not reproduce
>the issued you supplied.  I was able to evaluate and modify
>repeatedly and the correct values were maintained.

>If you feel this is still a problem resubmit your bug report
>with an code example, project\compiler settings, and steps
>to reproduce.  These details are necessary to duplicate the
>problem at hand.  If we can't reproduce the problem we can't
>forward it for further processing.

April 05, 1996
I have verifed the previously reported problem with the following code:

int foo(int i) {
  int a[4];

  a[0] = i;
  a[1] = i;
  a[2] = i;
  a[3] = i;
  a[0] *= 2;
  a[1] *= 2;
  a[2] *= 2;
  a[3] *= 2;
  return a[3];
}

int PASCAL WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance,  LPSTR
lpszCmdLine, int cmdShow) {
    int k;
  k = foo(1040);
}

Steps to create the problem are:
  1) Run to line "a[1] = i;" in foo.
  2) Watch a, i.
  3) Inspect i;
  4) In inspect window, change value of i to 33;
  5) Single step to line  a[0] *= 2;
  6) Watch window and inspect window show i =1040 again.

Note that a[1], a[2], and a[3] are assigned the proper value of 33, but as soon
as i is no longer used it is reported incorrectly.

>April 05, 1996
>Thank you for updating the information concerning your bug report.
> The new information has been accepted by the Techinical [sic] Support
> department for further investigation.
>
> Your reference number for this bug is still:   R002006
Note: See My bug report #13.1


*******************************************************************************************
My bug report #7, Borland bug report R001914
March 31, 1996
Subject:  Expressions containing pointer subtraction display ?? in watch
window.
April 10, 1996
>Thank you for taking the time to submit a bug report.
>Unfortunately, I have been unable to duplicate the
>problem you described.  If you wish to pursue this
>matter further, I suggest you resubmit the bug report with a
>small code sample, and refer to bug #1914.

April 12, 1996
The code sample below illustrates the problem of pointer subtraction in the
watch window displaying ??.
#include <windows.h>
int a[2], i, *x, *y;

#pragma argsused
int PASCAL WinMain(HINSTANCE hCurInstance, HINSTANCE hPrevInstance, LPSTR
lpCmdLine, int nCmdShow) {
  x = a;
  y = a+1;
  i = y-x;
  return i;
}

Watch i and y-x.
Single step to the "return i" instruction.
i = 1 as it should, but the expression y-x displays ?? in the watch window.

>April 16, 1996
>Thank you for taking the time to submit a bug report.
>Unfortunately, I was unable to duplicate the problem
>you described.  You may wish to resubmit this bug
>report with a brief code sample which illustrates the
>problem.  If you do, be sure to reference Bug number
>1914.
>Sincerely,
>Nick Rodin
>Borland Technical Support

<NOTE: Sound familiar?>

April 17, 1996
OK, I'll try this one more time.  Note that my last report on this bug INCLUDED
sample code to reproduce the problem. You then responded with
 "Unfortunately, I was unable to duplicate the problem
  you described.  You may wish to resubmit this bug
  report with a brief code sample which illustrates the
  problem."

I am still unable to correctly watch ANY expression involving the subtraction
of pointers.
Here is yet ANOTHER example:

Open the examples\windows\hdump project.
Single step to start the program.
Enter the following valid C expression in the watch window:
   &szDefExt[1] - &szDefExt[0]

On my system, the watch window displays ??.       What does yours
say?????????????

PLEASE DO NOT RESPOND THIS TIME BY TELLING ME THAT I MAY RESUBMIT THIS BUG
REPORT WITH A BRIEF CODE SAMPLE.  I HAVE DONE THIS TWICE.  PLEASE JUST TELL ME
WHAT RESULT YOU GET FOR THE ABOVE WATCH EXPRESSION SO THAT I KNOW IF ANYONE IS
ACTUALLY READING THIS.

>April 18, 1996
>Thank you for updating the information concerning your bug report.
> The new information has been accepted by the Techinical Support
> department for further investigation.
>
> Your reference number for this bug is still:   R001914



*******************************************************************************************
My bug report #8, Borland bug report R001908
March 31, 1996
Subject:  Font size in Stack and Inspect window.

When font size is changed for open window it appears properly.  Next time the
stack or inspect windows are opened, font size appears 5 points smaller.

>April 03, 1996
>This is a bug and will be rectified as soon as possible.  Thank you
>very much for bringing our attention to this matter.



*******************************************************************************************
My bug report #9, Borland bug report R001909
March 31, 1996
Subject:  Clicking the close button on the message window  (or otherwise
"activating" the message window such as by hitting Cntl-F6) moves the editor
window to the line of the message.

This is very annoying.  It should require at least a click on a line in the
message window to do this, if not a double-click.

Similarly, if message window is active when a breakpoint is hit, clicking
anywhere in the editor window moves the editor window to the line of the
warning instead of the execution point.

>April 01, 1996
>We have succeeded in reproducing the problem you describe and
>have forwarded the report to our development team for further processing.



*******************************************************************************************
My bug report #11, Borland bug report R001911
March 31, 1996
Subject:  Help bugs:

1) Cutting and pasting from help (e.g. unions example):
    a) spaces are pasted as 0xA0 instead of 0x20
    b) stupid Borland C++ 5.0 Programmer's Guide message appended.  I could
live without this annoyance.

2) Help on WaveProc topic: if you click on WaveInOpen (or any low level audio
function) you get "You have requested information from the Microsoft Win32
DeveloperÆs Reference. One or more of these  help files is not available on
your system."

3) Commenting out the line ":INDEX Microsoft Windows SDK Reference = WIN32.HLP"
in openhelp.cnt as mentioned in readme.txt prevents F1 help for windows SDK
functions to work.  This should have at least been clearly stated, as it quite
a serious limitation.  Sacrificing help on windows functions to have an index
of the other help topics is a lousy tradeoff to ask a programmer to make.  This
problem requires a better fix.

4) Message box frequently appears which says only "Windows Help" with OK and
Cancel buttons.


<Note: no response received>


*******************************************************************************************
My bug report #12, Borland bug report R001912
March 31, 1996
Subject:  The following are not bugs but are usability issues.

1) Watch Window and Inspect window should not wrap around from top to bottom
with up and down keys.
   This is confusing and inconsistant with other windows.
   The home and End keys work just fine get to the top and bottom lines.

2) Single step takes 2 seconds on a P-90.  This is with only six variables
displayed in the watch window.  It speeds along at about 1 second per step with
no variables watched.  Is there a reason this should take longer than 10 mS?
Why is it that changing the IDE to 32 bit code made it much slower?

Note: This was after I increased my RAM to 24 MB.  With 16 MB, I considered the
IDE unusable, as single stepping took 20 to 30 seconds!


<Note: no response received>


*******************************************************************************************
My bug report #13, Borland bug report R001913
March 31, 1996
Subject:  Wouldn't it be nice:

    If "variable declared but never used" messages pointed to the variable's
declaration line,  not to the end of the function.

    If in CPU window you could right click on variables to inspect, watch, etc.


    If Inspect windows didn't automatically close whenever you reset the
program.

    If you didn't have to now open and close *two* windows to change any of the
common watch options which are now considered "advanced".  What were you guys
thinking?

    If "evaluation tips" updated immediately after a single step instead of
every 4 seconds?

    If software companies would try Beta testing their products before shrink
wrapping them?

<Note: no response received>



*******************************************************************************************
My bug report: 13.1
April 5, 1996
Subject : Bug with bug reports

I have just emailed 3 bug reports, each providing more information for a
previously submited bug report and referencing the number of that bug report
under section 9 of TI 1226.
The referenced bug reports were R001916, R001907, and R001927.

For each of these submissions I received a confirmation from your autoresponder
which contained the following:

"Your reference number for this bug is still:   R002006"

I think your bug reporting system is buggy.


>April 05, 1996
>Thank you for submitting a bug report to Borland. In order to help
> expedite the turn-around time on the bug reports we receive, we have
> automated the process.  We have determined that your bug report isn't
> in the proper bug report format.  Please use the TI1226 form (refer to
> TI1171 for instructions on filling out TI1226) to resubmit your bug
> report.  Both TI1226 and TI1171 can be obtained as TI1226.ZIP and
> TI1171.ZIP from the following locations:
>Internet FTP: (anonymous login)
>   ftp.borland.com/pub/techinfo/techdocs/language/c++/bcpp/ti
>
> Compuserve Forum:
>   BCPPWIN LIB-2
>   BCPPDOS LIB-2
>
> Borland BBS:
>   408-431-5096 (14.4k bps N81)
>
> Once you have answered the questions by filling out the TI1226 bug
> report form please send it to us via:
>
>  1. Internet EMAIL (sent to 75300.1653@@compuserve.com)
>  2. Compuserve mail (sent to 75300,1653)
>  3. Our local BBS (408-431-5096.  From the main menu select 'Files',
>     then select 'Upload' and specify area 122 for Language Bug Reports.
>     Pick a filename with the extension .BUG)
>
> -- Borland AutoTech


I was trying to notify you of a bug in your bug reporting system.  Your
autoresponder responded with yet another bug by sending me a message that I
need to use TI 1226 which can be found at
ftp.borland.com/pub/techinfo/techdocs/language/c++/bcpp/ti.

This is not correct.  The correct URL is:
ftp.borland.com/pub/techinfo/techdocs/language/cpp/bcpp/ti.

PS.  I have never seen a company screw up so many things so badly as Borland
has done with their C products and their bug reporting process.  I could write
a book, but no one would beleive it.


<Note: original bug report 13.1 resubmitted April 5, 1996 on TI1226 with no
response from Borland.  Arrrrgghh!>



*******************************************************************************************
My bug report #14, Borland bug report TBD
April 18, 1996
Subject:  Evaluation Tips script stops working:

1) Load Evaluation Tips script.
2) Single step.
3) Make a change to source code.
4) Hit Single step.  A dialog appears asking "Make the modified code?"
5) Respond by hitting "No" (Hitting "Yes" also causes the problem).
6) Single step.  Evaluation Tips no longer function and cannot be restored
without closing and reopening the IDE.