www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Question Regarding Fixing A Bug [Issue 23654]

reply kdevel <kdevel vogtner.de> writes:
In January 2023 I reported a bug in [execv_: toAStringz: memory 
corruption](https://issues.dlang.org/show_bug.cgi?id=23654), 
provided a fix and had a discussion [std.process: memory 
allocation with malloc in 
execv_](https://forum.dlang.org/post/oeyisfsffcijzwlwcpaj forum.dlang.org).

That bug seems to be in the codebase for seven years, represents 
a regression, and has priority 1. But it is still unpatched.

If I read the discussion correctly Steven asked for disabling the 
GC before the arguments of the call to `exec*` are set up. I 
strongly disliked that because disabling the GC is useless in 
most of the cases (invocations) because the collector does not 
run anyway. There are no "cycles saved" by disabling the 
collector.

Disabling the GC and re-enabling it in a scope guard 
unnecessarily complicates the code.

If there is a misconception on my side, e.g. "most new allocation 
do not invoke the collector", I would appreciate a pointer to 
some documentation.
Aug 26 2023
parent reply Walter Bright <newshound2 digitalmars.com> writes:
On 8/26/2023 1:22 PM, kdevel wrote:

 In January 2023 I reported a bug in [execv_: toAStringz: memory 
 corruption](https://issues.dlang.org/show_bug.cgi?id=23654), provided a fix
and 
 had a discussion [std.process: memory allocation with malloc in 
 execv_](https://forum.dlang.org/post/oeyisfsffcijzwlwcpaj forum.dlang.org).
 
 That bug seems to be in the codebase for seven years, represents a regression, 
 and has priority 1. But it is still unpatched.
 
 If I read the discussion correctly Steven asked for disabling the GC before
the 
 arguments of the call to `exec*` are set up. I strongly disliked that because 
 disabling the GC is useless in most of the cases (invocations) because the 
 collector does not run anyway. There are no "cycles saved" by disabling the 
 collector.
 
 Disabling the GC and re-enabling it in a scope guard unnecessarily complicates 
 the code.
 
 If there is a misconception on my side, e.g. "most new allocation do not
invoke 
 the collector", I would appreciate a pointer to some documentation.



Can you please turn your patch into a Pull Request on github:

https://github.com/dlang/phobos

? Probably nobody had noticed the patch in bugzilla.
Aug 26 2023
parent reply kdevel <kdevel vogtner.de> writes:
On Sunday, 27 August 2023 at 02:41:43 UTC, Walter Bright wrote:

 [...]
 Can you please turn your patch into a Pull Request on github:

 https://github.com/dlang/phobos


Can you do that on behalf of me? I have no GH account.
Aug 26 2023
parent kdevel <kdevel vogtner.de> writes:
On Sunday, 27 August 2023 at 06:12:55 UTC, kdevel wrote:

 On Sunday, 27 August 2023 at 02:41:43 UTC, Walter Bright wrote:

 [...]
 Can you please turn your patch into a Pull Request on github:

 https://github.com/dlang/phobos


 Can you do that on behalf of me? I have no GH account.


The patch issue-23654.v0.patch is attached to the bugzilla ticket.
Oct 22 2023