Horton
Menu

Thread

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Use pg_malloc_object() and pg_alloc_array() variants in frontend code

  2. fe_utils: Sprinkle some pg_malloc_object() and pg_malloc_array()

  1. Use pg_malloc macros in src/fe_utils

    Henrik TJ <henrik@0x48.dk> — 2026-02-18T15:05:09Z

    Hi

Inspired by [1], I took a stab at converting src/fe_utils to the new 
pg_malloc macros.

Some mostly useless metrics: 31/34 converted, 7 casts removed.

meson test looks good for me.

1. https://www.postgresql.org/message-id/flat/CAHut%2BPvpGPDLhkHAoxw_g3jdrYxA1m16a8uagbgH3TGWSKtXNQ%40mail.gmail.com


   best regards, Henrik

  2. Re: Use pg_malloc macros in src/fe_utils

    Andreas Karlsson <andreas@proxel.se> — 2026-02-23T02:17:52Z

    On 2/18/26 4:05 PM, Henrik TJ wrote:
> Inspired by [1], I took a stab at converting src/fe_utils to the new 
> pg_malloc macros.
Thanks for the patch!

Looks like a nice change but why not just fix all instances of it in one 
swoop? It cannot be that many as there are 166 calls to pg_malloc() and 
62 calls to pg_malloc0() after your patch that need to be looked at.

-- 
Andreas Karlsson
Percona

  3. Re: Use pg_malloc macros in src/fe_utils

    Michael Paquier <michael@paquier.xyz> — 2026-02-24T03:36:50Z

    On Mon, Feb 23, 2026 at 03:17:52AM +0100, Andreas Karlsson wrote:
> Looks like a nice change but why not just fix all instances of it in one
> swoop? It cannot be that many as there are 166 calls to pg_malloc() and 62
> calls to pg_malloc0() after your patch that need to be looked at.

FWIW, I don't really mind if these changes are proposed gradually, and
this looked fine enough on its own.  So applied.
--
Michael

  4. Re: Use pg_malloc macros in src/fe_utils

    Andreas Karlsson <andreas@proxel.se> — 2026-02-27T01:15:46Z

    On 2/24/26 4:36 AM, Michael Paquier wrote:
> On Mon, Feb 23, 2026 at 03:17:52AM +0100, Andreas Karlsson wrote:
>> Looks like a nice change but why not just fix all instances of it in one
>> swoop? It cannot be that many as there are 166 calls to pg_malloc() and 62
>> calls to pg_malloc0() after your patch that need to be looked at.
> 
> FWIW, I don't really mind if these changes are proposed gradually, and
> this looked fine enough on its own.  So applied.


Fair, here is a patch which should handle all uses in the frontend code 
so we follow this pattern consistently to encourage new code to use 
these macros.

When doing this I found two things which I am ot sure what the cleanest 
way to handle would be so I broke them out into separate patches.

1. What should we do about when we allocate a an array of characters? 
Would it make sense to use pg_array_alloc() or would that jsut be silly? 
For example:

-pad = (char *) pg_malloc(l + 1);
+pad = pg_malloc_array(char, l + 1);

2. I found a small and harmless thinko. The buffer in verify_tar_file() 
is actually a char * but for some reason the code did the following:

buffer = pg_malloc(READ_CHUNK_SIZE * sizeof(uint8));

What should we do about it? Just skip the "sizof(uint8)"?

Andreas

  5. Re: Use pg_malloc macros in src/fe_utils

    Michael Paquier <michael@paquier.xyz> — 2026-02-27T10:00:00Z

    On Fri, Feb 27, 2026 at 02:15:46AM +0100, Andreas Karlsson wrote:
> 1. What should we do about when we allocate a an array of characters? Would
> it make sense to use pg_array_alloc() or would that jsut be silly? For
> example:
> 
> -pad = (char *) pg_malloc(l + 1);
> +pad = pg_malloc_array(char, l + 1);

I can see that tar_get_file_name() has been changed in 0001, which is
fine, so I have merged the change from 0002 in
dir_get_file_name()@walmethods.c into 0001, for consistency.  I don't
really have a strong opinion about the rest of 0002, TBH. 

> 2. I found a small and harmless thinko. The buffer in verify_tar_file() is
> actually a char * but for some reason the code did the following:
> 
> buffer = pg_malloc(READ_CHUNK_SIZE * sizeof(uint8));
> 
> What should we do about it? Just skip the "sizof(uint8)"?

This one has already been discussed, see here:
https://www.postgresql.org/message-id/aUJ2zxgPCaVsVi2a@ip-10-97-1-34.eu-west-3.compute.internal
The story is a bit larger than this single allocation, as it impacts
the meaning of the surrounding routines with backup manifests.

And applied 0001 after double-checking it.  Thanks.
--
Michael

  6. Re: Use pg_malloc macros in src/fe_utils

    Andreas Karlsson <andreas@proxel.se> — 2026-03-03T03:44:41Z

    On 2/27/26 11:00 AM, Michael Paquier wrote:
> On Fri, Feb 27, 2026 at 02:15:46AM +0100, Andreas Karlsson wrote:
>> 1. What should we do about when we allocate a an array of characters? Would
>> it make sense to use pg_array_alloc() or would that jsut be silly? For
>> example:
>>
>> -pad = (char *) pg_malloc(l + 1);
>> +pad = pg_malloc_array(char, l + 1);
> 
> I can see that tar_get_file_name() has been changed in 0001, which is
> fine, so I have merged the change from 0002 in
> dir_get_file_name()@walmethods.c into 0001, for consistency.  I don't
> really have a strong opinion about the rest of 0002, TBH.

Then I think we should skip it. If someone else wants to fix it in the 
future they are free to do so.

>> 2. I found a small and harmless thinko. The buffer in verify_tar_file() is
>> actually a char * but for some reason the code did the following:
>>
>> buffer = pg_malloc(READ_CHUNK_SIZE * sizeof(uint8));
>>
>> What should we do about it? Just skip the "sizof(uint8)"?
> 
> This one has already been discussed, see here:
> https://www.postgresql.org/message-id/aUJ2zxgPCaVsVi2a@ip-10-97-1-34.eu-west-3.compute.internal
> The story is a bit larger than this single allocation, as it impacts
> the meaning of the surrounding routines with backup manifests.

Thanks for the link!

> And applied 0001 after double-checking it.  Thanks.

Thanks!

-- 
Andreas Karlsson
Percona