Thank you for the feedback, Dave! Given you reviewed the change set and did not identify any regressions, I just have merged it via Files-ct.189/System-ct.1297. :-)

Best,
Christoph

---
Sent from Squeak Inbox Talk

On 2022-01-31T12:38:01-05:00, lewis at mail.msen.com wrote:

> Hi Christoph,
> 
> Your changes in DoItFirstAfterStartup.1.cs look correct to me. I
> tested on Linux and the --cwd command line option still works as
> expected. So +1 for putting this in trunk. As an aside, I notice
> that the existing FileDirectory class>>startUp method is mis-categorized,
> we should move it to "system start up".
> 
> I am happy to do the merge if you like, but I'm not sure if I can do it today.
> 
> On Sun, Jan 30, 2022 at 09:02:22PM +0100, christoph.thiede at student.hpi.uni-potsdam.de wrote:
> > Hi Marcel, Hi Dave,
> > 
> > I don't see why you would expect this to work. The normal SourceFiles array already contains a writable pointer to the changes file, so of course it cannot be opened writable again. There can only be one writable pointer to a file at one time, at least on Windows. I am more surprised that this rule does not apply on Linux/macOS.
> > 
> 
> The restriction on opening multiple writeable pointers to a file
> is a Windows design constraint that does not exist on other
> platforms. Like many such things, it is either a bug or a feature
> depending on your point of view ;-)
> 
> 
> > On Windows, apparently we need to close the changes file before re-opening it. We can either implement this manually or apply my proposed patch instead, because even conceptually, I don't see a real sense in making #openSourceFiles itself idempotent. :-)
> > 
> 
> Your changes handle this is a good way. So yes, calling #closeSourcesFiles
> first would probably take care of the issue on Windows, but I prefer
> your solution.
> 
> > Dave, you were probably unable to reproduce this in a debugger because once you have replaced the SourceFiles with a read-only pointer and old pointer has been GC'ed, the issue does not occur any longer. There is no limit for parallel read-only pointers on a file.
> > 
> > Best,
> > Christoph
> > 
> 
> Ah, good. It makes perfect sense now :-)
> 
> Dave
> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20220131/7f55b6a1/attachment.html>