Tom F

Hi, Stefan, But, if I shipped my app without debug info, and I want to find an address in THAT build, how do I get to a breakpoint?

Hi, Uwe, Thanks for your response. That's where I got lost. I shipped my app with no debugging info. If I want to trace down an address that's accurate in that version, I'd compile with the same flags as my shipped version, but this time generate a map file, right? That's no problem. But, then, how do I get to any debug windows (like the CPU window) when I start the app in the IDE with debugging, since there are no settable breakpoints.

What do you mean by a "detailed statement," Stano?

You're welcome. It took a lot of time to write. Good luck on your project. I'm headed back to coding. See the earlier post by @Stefan Glienke about your #3 being less likely to fail in the future if you change the enum definition.

Mike, You asked me what I meant by digging "too deeply." Suppose you want to drive from New York City to San Francisco. It's about a 43-hour drive. Suppose your friend says, "Oh, I can make your trip shorter. When you get to Chicago, there's a very long stoplight you'll probably have to sit at for two minutes waiting for it to turn green. But, to avoid the slowdown, I know a great detour. It'll save you about a minute. Turn left when you get to Main Street, and then turn right at the next street, go two blocks, and then turn right again and then turn at the next left. This will take you around the block, avoiding the stoplight delay, and then rejoin your original route to continue on your way to San Francisco." Would you take the detour? The detour is faster than staying on the route with the long stoplight. But is it enough faster for you to worry about? Is it enough faster to bother taking? Is it enough faster to risk getting lost on the streets of Chicago? Is it enough faster to spend hours discussing on forums? How much faster should the detour be before you're willing to take it? How much risk can the detour add before you think to yourself, "I'm not going to take that detour." ??? In your code, you have no idea how long the trip is. You have no idea where the bottleneck slowdowns are. So, you have no idea if the proposed speedup will be significant. And people here in the forum are telling you that your detour (new code) may be riskier (i.e. more difficult to maintain.) Suppose your enum code is currently consuming a total of .0050 seconds every time a user clicks a certain button. Suppose you spend hours writing several versions of alternate code. Suppose your new code speeds up enum strings so that they're now TWICE AS FAST, so that processing enums only takes .0025 seconds. Was it worth your time to work on this? What many of us have been saying to you here and elsewhere is that the revisions (detours) that you're suggesting to speed up your code are insignificant and may bring risk. You need to learn how to put your efforts where they will have the biggest impact on improving your code. The first rule is WRITE THE CLEANEST, MOST MAINTAINABLE CODE YOU CAN. Then MEASURE your application with a performance monitor/profiler to identify places where it's slow. A profiler will help you do this by showing you exactly where your code is spending its time. It's likely that the amount of time your current code is spending on enum strings is infinitesimal compared to everything else your code is doing. So even if you speed up the enum code, the speedup is likely to have no impact on your users.

If you're refactoring for maintainability, then I suggest you don't waste your time digging too deeply into speed benchmarks and certainly not into the assembly language.

If you haven't benchmarked your application to see how much time this code consumes in the context of your application, you have no way of knowing what "significantly slower" means. I'd bet that you could slow down this function by a factor of 1000 and it would NOT noticeably impact your users' experience or computer resources. You're wasting your time. Why have you chosen to waste your time working on something that doesn't matter? As Glienke says, "Version3 is the best because it's maintainable and gives compile error if anyone decides to add a new value..."

Hi, Mike, No, this wasn't posted in the wrong thread. You said you were refactoring. There are two reasons to refactor code that I'm aware of: 1. Make it easier to maintain. 2. Significantly improve the speed of the software. To answer your question about which approach is better, we need to know of the above two are you doing? Since you've provided speed benchmarks, I assume that it's #2 -- you're trying to significantly speed up your software. I suspect that you are prematurely optimizing your code for speed before you know where the bottlenecks are. Have you used a profiler to determine where to best focus on speeding up your application? Even if you made this code infinitely fast, would you or your users be able to notice the difference? If your answer is "yes" – then we'd love to hear about this novel application. If your answer is "no" – then you're wasting your time.

How did you decide that this was code that you wanted to optimize? Have you profiled your application to see if this conversion is a significant bottleneck? These aren't rhetorical questions. I am genuinely interested in your answers. I hope you'll answer them here in a posting.

I'm looking to capture usage data. AppAnalytics was formerly available, including access to a server and reports. Emb still makes the code available it seems, including server code. But apparently there is not subscription access to a server for actually collecting data, unless we provide the server. I suspect that's way over my technical skills. So, has anyone have a suggestion on similar usage metering similar tool with a Delph app (Sydney)? TIA

I'm about to update to 10.4, having waited while EMB QA once again failed and we had to wait for them to remove bugs that should never have shipped. (Yes, I'm bitter about that, but that's a discussion to have another day.) Rather than continuing to suffer from EBM releases that are defective, I'm going to for the first time develop on VMWare VM. This will of course allow me to easily roll back to a previous release. One reason I haven't done this in the past is that I have never been clear about how much I should install in the VM itself, and how much should remain on my host machine. If I put my source code (which is regularly backed up from the host to a repository) in the VM, then it seems that so many of my other tools also have to be in the VM, like my grep tools, and other utilities (like Beyond Compare, CI tools), etc. I don't know the best place to draw the line on what goes where. Perhaps I should leave my source on the host drive, accessing it from the VM? This is less than ideal, because my archives would have two pieces (the VM, and the source on C:). And it also means that I'll have to remove all of the hardwired C:\ paths in my scripts and other tools. To people using VMs for product development: how much do you install in your VM? Does your VM access your host drive for anything? Any suggestions?

Your workaround above did not work for me. The only work-around I've found is to restart the IDE, select "None" for Code insight manager , then open the .pas file that was causing the crash. If I want, I can then re-enable LSP and it works fine in all files, including the one that was causing difficulties. The above doesn't always work for me but usually playing around with the Code Insight manager setting for a few times after a few crashes finally lets me load the file.

Pat, loading an extra IDE doesn't solve this problem, which isn't a "disappearing act." The application bde.exe actually terminates and Windows generates a crash report. Dave, I've got 8 of 16 gigs of memory free, so a shortage of memory is not the problem. Thanks for the good suggestion though. I hadn't thought of that... I've discovered through a lot of trial and error that when I encounter a unit that when opening a file crashes the IDE, if I restart the IDE and toggle CodeInsight between LSP, Classic, and None, (sometimes a few times) I can usually load the problem .pas file. I haven't found any predictable sequence of toggling. Sometimes (but not always) I can load the .pas file into the IDE without a crash, regardless of whether CodeInsight is configured to LSP, classic, or None... 😞 This toggling procedure is not a fix, of course. It may simply be that there's some uninitialized memory that the above workaround leaves in a better state. The toggling procedure isn't even a reliable workaround since it apparently has to be done after the crash. All of this means I'm probably going to leave all my large forms and frames open in the IDE and not shut down the IDE for weeks at a time in an effort to avoid the problem. But at least I'm not stopped dead in my tracks. I also had a graceful crash in the IDE occur while I was testing this. I grabbed the call stack and other recommended files (like the DxDiag output) and appended them to the original RSP https://quality.embarcadero.com/browse/RSP-34109.

A quick note to anyone looking at this thread. You probably would have figured this out on your own, but, looking at the Task Manager, it appears that when you re-select the LSP on the Options screen, you have to close and re-open any open project for the IDE to re-launch the LSP.

Darn, I was afraid that's where you were referring to. Unfortunately, I don't have any of the git entries filled in. 😞 So that's not where my problem is headed. What bothers me is that the crash is so hard. There's no message or anything. The IDE just disappears. Most programs pop up a message box with exception errors. The Windows error log shows the crash, so perhaps somehow the IDE or my machine are now configured in some way to not show any crash messages. Or perhaps this particular crash is just too deep. Although the problem has been intermittent, at the current time I have a reliable way to make it crash. i.e. There is one particular.pas file that crashes every time I try to load it. I am making a VM backup of my development environment and noted in a comment in the RSP, hoping that EMB will ask for me to provide it.

Uwe: Thanks for the suggestion. I still get intermittent lockups when opening .pas files, but so far see no pattern. How does one disable the internal Git client?

Hi, Vandrovnik, Thanks for suggesting the command-line switches. I hid the splash screen and could see the new progress meter displaying the loading of my project's .pas source into the editor before it crashed. So I closed the IDE, and hid my source in a different folder, and was able to start the IDE. I then moved my source back to its original folder and all is fine for now, including being able to start the IDE. Somehow, the loading of one particular source file was causing a crash. However, the file loads fine now. I've turned off "Reopen last opened project when starting RAD Studio" I doubt the above steps fixed the problem. I expect the IDE will continue to intermittently crash, but at least I'll be able to restart it.

That's a reasonable suggestion, Vandrovnik. Except I can't uninstall any packages, because I can't run the IDE. Does uninstalling RAD Studio uninstall everything or are their folders and registry entries I have to clean. This is an intermittent bug, so I can't just "try again" and hope it works. I need a development environment that is stable.

Thanks, Remy (and to the other answers). Your approach is the one that appeals to me most. I'm glad it's working for you. I'll give it a try.

In the TDbGrid as shown, when the line below is executed: dataset1.RecNo = 7; Delphi correctly selects the row, but then scrolls the grid down a few rows vertically so that row 7 is vertically centered in the grid even though it was previously visible. Is there a way to disable this unconventional behavior?

I'm the OP on this thread. This post provides a solution to my issue: https://stackoverflow.com/questions/2995662/delphi-restore-actual-row-in-dbgrid/2996360#2996360.

In VCL, how to make a fsStayOnTop form the active form? I have a main form and a non-modal subform. When the Main form is active, if the user enters a specific key associated with a shortcut on the non-modal form, my main form calls a method in the non-modal form. But, how can I make the non-modal form the active form when it receives this call? I've tried to assign Screen.ActiveForm and SuForm.Active, but both are read-only.

TForm.SetFocus worked from the subform. Thank you, Anders!

In an app, some of our VCL Forms have FormStyle := fsStayOnTop assigned in the Object Inspector. Occasionally we temporarily change it to fsNormal when we're popping up a system font or color dialog box. Sometimes, (but so far unpredictably) including sometimes when Alt-tabbing back to our app, our on-top window is not visible. Sometimes the user can make it visible again with some combination of Alt-Tabs, and minimizing then un-minimizing the app on the Taskbar. As I understand it, this isn't unusual, and that under Windows, Microsoft doesn't guarantee that fsStayOnTop is reliable, given various contention issues, etc. I'm looking for a way to reduce the frequency of this problem in our app. What is the current "best practice" is in this area? There are a smattering of posts on the web on this topic, some go back decades. One of the posts suggests overriding CreatParams as shown below. Rather than being a total script-kiddie, is the code below a reasonable approach, or is it just the equivalent of setting FormStyle := fsStayOnTop? procedure TForm1.CreateParams(var Params: TCreateParams); begin inherited CreateParams(Params); with Params do begin Params.ExStyle := ExStyle or WS_EX_TOPMOST; end; end;

tl;dr: How can I configure a column in the TDbGrid (or TRzDbGrid if it's easier) so that when the user clicks in a cell, I get control to set the cell's background color? Details: I am using a TDbGrid that has a column for the user to select a color that is then used elsewhere in the application. There is no text in the cells in the column, just a color. In DrawColumnCell, for that column, I set Canvas.Brush.Color so that the cell displays a previously selected color. This column has ButtonStyle := cbsEllipsis; When the user clicks the ellipsis, I popup a TColorDialog for the user to choose a color. The above is all working fine, except when the user clicks in the cell (before clicking the ellipsis) the Brush color goes back to white. (See attached image.) I don't want to auto-open the color picker. I believe setting the editor color involves cracking the InplaceEditor but I haven't been able to find enough info online to figure out how to do that.